Agency Inventory AI Usage - Sheet1.tsv

Agency Name of Inventory Item Description of Inventory Item Primary Type of AI Purpose of AI Length of Usage Does it directly impact the public? Vendor System Other Notes 
USDA Agricultural Research Service - 4% Repair Dashboard The model reviews the descriptions of expenses tagged to repairs and maintenance and classifies expenses as "repair" or "not repair" based on keywords in context. Natural Language Processing Classification or Labeling Unknown No impact 
USDA Agricultural Research Service - Project Mapping Term analysis and clustering enables program leaders to find synergies and patterns across ARS research program portfolios. Natural Language Processing Project Management Unknown No impact 
USDA Agricultural Research Services - NAL Automated Indexing Uses machine learning for indexing of publication abstracts and project proposals using terms from USDA National Agricultural Library Thesaurus Machine Learning (Type Unknown) Classification or Labeling Unknown No impact Yes (Cogito) 
USDA Forecasting Grasshopper Outbreaks in the Western United States using Machine Learning Tools Integrate historic grasshopper survey data and grasshopper biology with environmental covariates (e.g., climate, soil, and topography) to generate grasshopper outbreaks forecasts for the western U.S. Maximum Entropy Model (MaxEnt) Forecasting & Prediction Unknown No impact 
USDA Agricultural Research Services - Facial Recognition Facial recognition as one of several factors for access to secure areas of a facility Facial Recognition Security Unknown Direct impact 
USDA Economic Research Service - Coleridge Initiative The purpose of this project is the use AI tools to understand how publicly funded data and evidence are used to serve science and society. Natural Language Processing Research (Other) Unknown No impact 
USDA Economic Research Service - Westat A competition to find automated, yet effective, ways of linking USDA nutrition information to 750K food items in a proprietary data set of food purchases and acquisitions Natural Language Processing Organization & Efficiency Unknown Indirect impact 
USDA Farm Production and Conservation - Land Change Analysis Tool Employ learning classifier to produce high resolution land cover maps from aerial and/or satellite imagery and publish results through publicly available Image service. Training data is generated from a custom-built web application. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Unknown No impact LCAT has mapped over 600 million acres and have generated over 700 thousand training samples. 
USDA Food and Nutrition Services - Retailer Receipt Analysis Use OCR on a sample of FNS receipt and invoice data; consultants will use this data to see how existing manual process can be automated, saving staff time, ensuring accurate review, and detecting difficult patterns. Will pave the way for a review system that (1) has an automated workflow and learns from analyst feedback (2) can incorporate known SNAP fraud patterns, look for new patterns, and visualize alerts on these patterns on retailer invoices and receipts. Optical Character Recognition (or Text Extraction) Organization & Efficiency Short-term project or study Indirect impact 
USDA Forest Service - Ecosystem Management Decision Support System (EMDS) EMDS is a spatial decision support system for landscape analysis and planning that runs as a component of ArcGIS and QGIS. Machine Learning (Type Unknown) Mapping Ongoing project (time unknown) No impact Users develop applications for their specific problem that may use any combination of four AI engines for 1) logic processing, 2) multi-criteria decision analysis, 3) Bayesian networks, and Prolog-based decision trees. 
USDA Forest Service - Wildland Urban Interface - Mapping Wildfire Loss Uses machine learning to identify buildings, building loss, and defensible space around buildings before and after a wildfire event in wildland-urban interface settings. Neural Networks Mapping Ongoing project (time unknown) Indirect impact Also uses image-based classification 
USDA Forest Service - National Land Cover Database (NLCD) Tree Canopy Cover Mapping Responsible for producing maps with consistent spatial resolution. The forest structure maps are generated using over 60,000 training plots with a probabilistic sample design to train statistical machine learning models to classify continuous tree canopy cover. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Unknown No impact 
USDA Forest Service - BigMAP project The project uses machine learning, along with features derived from dense time series of Landsat imagery as well as climatic and topographic data, to impute attributes from national forest inventory database to produce raster maps of U.S. forest resources. Machine Learning (Type Unknown) Mapping Unknown No impact 
USDA Forest Service - DISTRIB-II: Habitat Suitability of Eastern United States Trees Habitat suitability is modeled for 125 eastern United States trees species under 1981-2010 climate conditions and 8 projected future conditions (2070-2099). Unclear Forecasting & Prediction Short-term project or study Indirect impact The AI provides insight into options for managing eastern U.S. forests. 
USDA Forest Service - CLT Knowledge Database The CLT knowledge database catalogs cross-laminated timber information in an interface that helps users find relevant information. The information system uses data aggregator bots that search the internet for relevant information. These bots search for hundreds of keywords and use machine learning to determine if what is found is relevant. Machine Learning (Type Unknown) Research (Other) Ongoing project (time unknown) No impact As of 2/24/2022, the CLT knowledge database has cataloged >3,600 publications on various aspects of CLT. Manufacturers, researchers, design professionals, code officials, government agencies, and other stakeholders directly benefit from the tool, thereby supporting the increasing use of mass timber, which benefits forest health by increasing the economic value of forests. 
USDA Forest Service - RMRS Raster Utility RMRS Raster Utility is a .NET object-oriented library that simplifies data acquisition, raster sampling, and statistical and spatial modeling while reducing the processing time and storage space associated with raster analysis. Machine Learning (Type Unknown) Organization & Efficiency Ongoing project (time unknown) No impact 
USDA Forest Service - TreeMap 2016 TreeMap 2016 provides a tree-level model of the forests of the conterminous United States. It matches forest plot data from Forest Inventory and Analysis (FIA) to a 30x30 meter (m) grid. TreeMap 2016 is being used in both the private and public sectors for projects including fuel treatment planning, snag hazard mapping, and estimation of terrestrial carbon resources. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Ongoing project (time unknown) Indirect impact A random forests machine-learning algorithm was used to impute the forest plot data to a set of target rasters provided by Landscape Fire and Resource Management Planning Tools (LANDFIRE). 
USDA Forest Service - Landscape Change Monitoring System (LCMS) National Landsat/sentinel remote sensing-based data produced by the USDA Forest Service for mapping and monitoring changes related to vegetation canopy cover, as well as land cover and land use. Unclear Mapping Ongoing project (time unknown) No impact 
USDA Forest Service - Forest Health Detection Monitoring Machine learning models are used to (1) upscale training data that was collected from both the field and high-resolution imagery to map and monitor stages of forest mortality and defoliation, and (2) to post-process raster outputs to vector polygons. Machine Learning (Type Unknown) Mapping Unknown No impact 
USDA Forest Service - Land Cover Data Development Apply supervised classification methods with remotely sensed satellite aerial imagery and other landscape variables (e.g., digital elevation derivations, soils data, geology data, etc.) for labelling segments or pixels with land attributes for a landscape/study area. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Classification or Labeling Unknown No impact 
USDA National Agricultural Statistics Service - Cropland Data Layer Interpret readings from satellite-based sensors and classify the type of crop or activity that falls in each 30 square meter pixel on the ground. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production more than a year) No impact The CDL has been produced for national coverage since 2008. 
USDA National Agricultural Statistics Service - List Frame Deadwood Identification The deadwood model produces a propensity score representing a relative likelihood of a farm operation being out of business. Common tree splits were identified using the model and combined with expert knowledge to develop a recurring process for deadwood clean up. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Unknown Indirect impact 
USDA National Institute for Food and Agriculture - Climate Change Classification NLP The model classifies NIFA funded projects as climate change related or not climate related through natural language processing techniques Natural Language Processing Classification or Labeling Unknown No impact 
USDA Office of Safety, Security, and Protection - Video Surveillance System The Video Surveillance System shall control multiple sources of video surveillance subsystems to collect, manage, and present video clearly and concisely Facial Recognition Security Unknown Direct impact 
USDA Office of the Chief Information Officer - Acquisition Approval Request Compliance Tool NLP model developed to utilize the text in procurement header and line descriptions within USDA's Integrated Acquisition System to determine the likelihood that an award is IT-related, and therefore might require an AAR. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
USDA Operational water supply forecasting for western US rivers The USDA National Water and Climate Center operates the largest forecast system of spring-summer river flow volumes. The NWCC recently developed a next-generation prototype for generating such operational water supply forecasts, the multi-model machine-learning metasystem, which integrates a variety of AI and other data-science technologies carefully chosen or developed to satisfy specific user needs. Machine Learning (Type Unknown) Forecasting & Prediction Ongoing project (time unknown) Indirect impact 
USDED Federal Student Aid - Aidan Chat-Bot FSA's virtual assistant answers common financial aid questions and help customers get information about their federal aid on StudentAid.gov. Natural Language Processing Chat Bot Ongoing project (time unknown) Direct impact In just over two years, Aidan has interacted with over 2.6 million unique customers, resulting in more than 11 million user messages. 
USDOC International Trade Administration - B2B Matchmaking The system's algorithms and AI technology qualifies data and makes B2B matches with event participants according to their specific needs and available opportunities. Unclear Classification or Labeling Unknown No impact 
USDOC International Trade Administration - ChatBot Pilot Chatbot embedded into trade.gov to assist ITA clients with FAQs, locating information and content, suggesting events and services. Natural Language Processing Chat Bot Unknown Direct impact 
USDOC International Trade Administration - Consolidated Screening List The CSL search engine has “Fuzzy Name Search” capabilities, allowing a search without knowing the exact spelling of an entity’s name. In Fuzzy Name mode, the CSL returns a “score” for results that exactly or nearly match the searched name. Natural Language Processing Forecasting & Prediction Unknown No impact The Consolidated Screening List (CSL) is a list of parties for which the United States Government maintains restrictions on certain exports, reexports, or transfers of items. It consists of the consolidation of 13 export screening lists of the Departments of Commerce, State, and Treasury. 
USDOC International Trade Administration - AD/CVD Self Initiation The ADCVD program investigates allegations of dumping and/or countervailing of duties. Investigations are initiated when a harmed US entity files a petition identifying the alleged offence and the specific harm inflicted. Self-Initiation will allow ITA to monitor trade patterns for this activity and preemptively initiate investigations by identifying harmed US entities. Unclear Classification or Labeling Unknown Indirect impact 
USDOC International Trade Administration - Market Diversification Toolkit A user enters what products they make and the markets they currently export to. The Market Diversification Tool applies a ML algorithm to identify and compare potential new export markets that should be considered. The tool brings together product-specific trade and tariff data and economy-level macroeconomic and governance data to provide a picture of which markets make sense for further market research. Machine Learning (Type Unknown) Research (Other) Unknown No impact 
USDOC NOAA - Fisheries Electronic Monitoring Image Library The Fisheries Electronic Monitoring Library (FEML) will be the central repository for electronic monitoring (EM) data related to marine life. Automated Image Processing Monitoring or Detection Planning or development stage No impact 
USDOC NOAA - Passive acoustic analysis using ML in Cook Inlet, AK Passive acoustic data is analyzed for detection of beluga whales and classification of the different signals emitted by these species. Neural Networks Classification or Labeling Ongoing project (time unknown) No impact Results are being used to inform seasonal distribution, habitat use, and impact from anthropogenic disturbance within Cook Inlet beluga critical habitat. The project is aimed to expand to other cetacean species as well as anthropogenic noise. 
USDOC NOAA - AI-based automation of acoustic detection of marine mammals Command line software which was developed in-house for model training, evaluation, and deployment of machine learning models for the purpose of marine mammal detection in passive acoustic data. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (time unknown) No impact It also includes annotation workflows for labeling and validation. 
USDOC NOAA - Developing automation to determine species and count using optical survey data in the Gulf of Mexico Focuses on optical survey collected in the Gulf of Mexico: 1) develops an image library of landed catch, 2) develops automated image processing (ML/DL) to identify and enumerate species from underwater imagery and 3) develops automated algorithms to process imagery in near real time and download information to central database. Automated Image Processing Monitoring or Detection Unknown No impact 
USDOC NOAA - Fast tracking the use of VIAME for automated identification of reef fish Compiling image libraries for use in creating automated detection and classification models for use in automating the annotation process for the SEAMAP Reef Fish Video survey of the Gulf of Mexico. VIAME models are performing well enough that we will incorporate automated analysis in video reads soon as part of a supervised annotation-qa/qc process. Automated Image Processing Monitoring or Detection Ongoing project (time unknown) No impact 
USDOC NOAA - A Hybrid Statistical-Dynamical System for the Seamless Prediction of Daily Extremes and Subseasonal to Seasonal Climate Variability Demonstrate the skill and suitability for operations of a statistical-dynamical prediction system that yields seamless probabilistic forecasts of daily extremes and sub seasonal-toseasonal temperature and precipitation. Machine Learning (Type Unknown) Forecasting & Prediction Unknown Indirect impact Recently demonstrated a Bayesian statistical method for post-processing seasonal forecasts of mean temperature and precipitation from the North American Multi-Model Ensemble (NMME). 
USDOC NOAA - FathomNet FathomNet provides much-needed training data (e.g., annotated, and localized imagery) for developing machine learning algorithms that will enable fast, sophisticated analysis of visual data. Machine Learning (Type Unknown) Research (Other) Ongoing project (time unknown) No impact 
USDOC NOAA - ANN to improve CFS T and P outlooks Using Artificial Neural Networks to Improve CFS Week 3-4 Precipitation and Temperature Forecasts Neural Networks Forecasting & Prediction Unknown Indirect impact 
USDOC NOAA - Drought outlooks by using ML techniques Drought outlooks by using ML techniques with NCEP models Machine Learning (Type Unknown) Forecasting & Prediction Ongoing project (time unknown) Indirect impact 
USDOC NOAA - EcoCast Operational tool that uses boosted regression trees to model the distribution of swordfish and bycatch species in the California Current. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Unknown No impact 
USDOC NOAA - Coastal Change Analysis Program (C-CAP) C-CAP embarked on operational high resolution land cover development effort that utilized geographic object-based image analysis and ML algorithms such as Random Forest to classify coastal land cover from 1m multispectral imagery. Automated Image Processing Classification or Labeling Ongoing project (in production more than a year) No impact 
USDOC NOAA - Deep learning algorithms to automate right whale photo id AI for right whale photo ID has expanded to include several algorithms to match right whales from different viewpoints (aerial, lateral) and body part (head, fluke, peduncle). Neural Networks Classification or Labeling Unknown No impact 
USDOC NOAA - NN Radiation Developing fast and accurate NN LW- and SW radiations for GFS and GEFS. Neural Networks Forecasting & Prediction Short-term project or study No impact 
USDOC NOAA - NN training software for the new generation of NCEP models Optimize NCEP EMC Training and Validation System for efficient handling of high spatial resolution model data produced by the new generation of NCEP's operational models. Neural Networks Organization & Efficiency Unknown No impact 
USDOC NOAA - Coral Reef Watch Offering the world's only global early-warning system of coral reef ecosystem physical environmental changes, CRW remotely monitors conditions that can cause coral bleaching, disease, and death; delivers information and early warnings in near real-time to our user community; and uses operational climate forecasts to provide outlooks of stressful environmental conditions at targeted reef locations worldwide. CRW products are primarily sea surface temperature (SST)-based but also incorporate light and ocean color, among other variables. Unclear Forecasting & Prediction Ongoing project (time unknown) No impact 
USDOC NOAA - Robotic microscopes and machine learning algorithms remotely and autonomously track lower trophic levels for improved ecosystem monitoring and assessment Deploy the Imaging Flow Cytobot on fixed (docks) and roving (aboard survey ships) platforms to autonomously monitor phytoplankton communities in aquaculture areas in Puget Sound and in the California Current System. Map the distribution and abundance of phytoplankton functional groups and their relative food value to support fisheries and aquaculture and describe their changes in relation to ocean and climate variability and change. Automated taxonomic identification of imaged phytoplankton uses a supervised machine learning approach (random forest algorithm). Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Unknown No impact 
USDOC NOAA - Edge AI survey payload development This is a nine camera (color, infrared, ultraviolet) payload controlled by dedicated on-board computers with GPUs. YOLO detection models run at a rate faster than image collection, allowing real-time processing of imagery as it comes off the cameras. Goals of effort are to reduce overall data burden and reduce the data processing timeline, expediting analysis and population assessment for arctic mammals. Automated Image Processing Classification or Labeling Unknown No impact 
USDOC NOAA - Ice seal detection and species classification in multispectral aerial imagery Refine detection and classification pipelines with the goal of reducing false positive rates (to < 50%) while maintaining > 90% accuracy and significantly reducing labor intensive, post survey review process. Automated Image Processing Monitoring or Detection Unknown No impact 
USDOC NOAA - First Guess Excessive Rainfall Outlook First guess for the WPC Excessive Rainfall Outlook - It is learned from the ERO with atmospheric variables. Machine Learning (Type Unknown) Forecasting & Prediction Unknown No impact 
USDOC NOAA - CoralNet Operational point annotation software for benthic photo quadrat annotation; development of classifiers allows for significantly reducing human annotation. Machine Learning (Type Unknown) Classification or Labeling Unknown No impact 
USDOC NOAA - Automated detection of hazardous low clouds in support of safe and efficient transportation Maintenance and sustainment project for the operational GOES-R fog/low stratus (FLS) products, routinely used by the NWS Aviation Weather Center and Weather Forecast Offices. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact The FLS products are derived from the combination of GOES-R satellite imagery and NWP data using machine learning. 
USDOC NOAA - The Development of ProbSevere v3 ProbSevere is a ML model that utilizes NWP, satellite, radar, and lightning data to nowcast severe wind, severe hail, and tornadoes. ProbSevere v3 utilizes additional data sets and improved machine learning techniques to improve upon the operational version of ProbSevere. Machine Learning (Type Unknown) Forecasting & Prediction Unknown Direct impact ProbSevere v3 utilizes additional data sets and improved machine learning techniques to improve upon the operational version of ProbSevere. 
USDOC NOAA - The VOLcanic Cloud Analysis Toolkit: System for detecting, tracking, characterizing, and forecasting hazardous volcanic events Consists of several AI powered satellite applications including: eruption detection, alerting, and volcanic cloud tracking. These applications are routinely utilized by Volcanic Ash Advisory Centers to issue volcanic ash advisories. Unclear Monitoring or Detection Unknown Direct impact 
USDOC NOAA - SUVI Thematic Maps The SUVI Thematic Maps product is a Level 2 data product that (presently) uses a machine learning classifier to generate a pixel-by-pixel map of important solar features digested from all six SUVI spectral channels. Machine Learning (Type Unknown) Mapping Unknown No impact 
USDOL Form Recognizer for Benefits Forms Custom machine learning model to extract data from complex forms to tag data entries to field headers. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production less than six months) Direct impact Machine learning model uses computer vision. 
USDOL Claims Document Processing Identifies if a physician’s note contains causal language by training custom natural language processing models Natural Language Processing Monitoring or Detection Planning or development stage Direct impact Natural language processing for (a) document classification and (b) sentence-level causal passage detection 
USDOL Website Chatbot Assistant The chatbot helps the end user with basic information about the program, information on who to contact, or seeking petition case status. Natural Language Processing Chat Bot Planning or development stage Direct impact 
USDOL Data Ingestion of Payroll Forms Custom machine learning model to extract data from complex forms to tag data entries to field headers. Natural Language Processing Classification or Labeling Planning or development stage Direct impact 
USDOL Hololens AI used by Inspectors to visually inspect high and unsafe areas from a safe location. Unclear Mapping Ongoing project (in production more than a year) No impact 
USDOL SOII Computer-Assisted Coding The Survey of Occupational Injuries and Illnesses (SOII) collects hundreds of thousands of narratives describing cases of work-related injury and illness annually. Autocoders assign classifications for worker occupation, nature of injury, part of body, event or exposure, source, and secondary source for each case. Natural Language Processing Classification or Labeling Ongoing project (in production more than a year) No impact Autocoders subsequently expanded and coded 85% of all SOII elements for reference year (RY) 2019. This gradual increase occurred by adapting the selection criterion based on careful monitoring of the processes; project also uses deep neural networks with character-level convolutional embeddings and Long-Short-Term-Memory recurrent layers 
USDOS Bureau of Global Public Affairs - CLIPSLAB GPA’s production media collection and analysis system that pulls data from half a dozen different open and commercial media clips services to give an up-to-date global picture of media coverage around the world. Network Analysis (ie Bayesian, or Social Network) Organization & Efficiency Ongoing project (time unknown) No impact 
USDOS Bureau of Global Public Affairs - Mission Press Digest A prototype system that collects and analyzes the daily media clips reports from about 70 different Embassy Public Affairs Sections. Organization & Efficiency Ongoing project (time unknown) No impact 
USDOS Bureau of Global Public Affairs - Digital Communications Database GPA’s production system for collecting, analyzing, and summarizing the global digital content footprint of the Department. Organization & Efficiency Ongoing project (time unknown) No impact 
USDOS Bureau of Global Public Affairs - Facebook Ad Test Optimization System GPA’s production system for testing potential messages at scale across segmented foreign sub-audiences to determine effective outreach to target audiences. Forecasting & Prediction Ongoing project (time unknown) Direct impact 
USDOS Bureau of Global Public Affairs - Global Audience Segmentation Framework GPA’s prototype framework for predicting how content or messages designed for one audience some place in the world will resonate with other audiences outside the United States. Forecasting & Prediction Ongoing project (time unknown) Direct impact 
USDOS Bureau of Global Public Affairs - Machine-Learning Assisted Measurement and Evaluation of Public Outreach A high-performing classifier capable of measuring the level of six different emotions that a text evokes, including time-series analysis, to help evaluate historical messages and predict successful future public messaging. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (time unknown) Direct impact 
USDOS Bureau of Global Public Affairs - GPA Tools and GPAIX AI-enabled analysis package for automating public outreach analysis. Unclear Research (Other) Ongoing project (time unknown) Direct impact 
USDOS Bureau of Political-Military Affairs - Pull Information from Unstructured Text Use natural language processing to extract information from document text to help summarize and allow for analysis more efficiently than manual methods. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact 
USDOS Bureau of Political-Military Affairs - K-Means Clustering Into Tiers Cluster countries into tiers based on data collected from open source and Bureau data using k-means clustering. Clustering (K-means, etc.) Classification or Labeling Ongoing project (time unknown) No impact 
USDOS Global Engagement Center - Disinformation Topic Modeling Text clustering and topic modeling of documents and social media to determine possible disinformation subjects and topics. Clustering (K-means, etc.) Mapping Ongoing project (time unknown) No impact 
USDOS Global Engagement Center - Deepfake Detector Classifies facial images as either being real (contains a real person’s face) or fake (synthetically generated face, a deepfake often created using Generative Adversarial Networks) to predict disinformation activities. Neural Networks Classification or Labeling Ongoing project (time unknown) No impact 
USDOS Global Engagement Center - Text Similarity Detection Identifies different texts that are identical or nearly identical by calculating cosine similarity between each pair of texts. Texts are then grouped if they share high cosine similarity and then available for analysts to review further. Natural Language Processing Research (Other) Ongoing project (time unknown) No impact 
USDOS Global Engagement Center - Image Clustering for Disinformation Detection Identifies similar images in order to analyze how images are used to spread and build traction with disinformation narratives. Automated Image Processing Research (Other) Ongoing project (time unknown) No impact 
USDOS Global Engagement Center - Louvain Community Detection Clusters nodes together into “communities” to detect clusters of accounts possibly spreading disinformation. Network Analysis (ie Bayesian, or Social Network) Research (Other) Ongoing project (time unknown) No impact 
USDOS Office of U.S. Foreign Assistance Resources - Foreign Assistance Appropriations Automates and streamlines the extraction of earmarks and directives from the annual appropriations bill to facilitate the Department’s adherence to congressional direction. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact 
USDOS Office of Management Strategy and Solutions - Department Cables Analytics Analysis of Department cables reporting to inform multiple areas of Department policy and operations. Natural Language Processing Project Management Ongoing project (time unknown) No impact 
USDOS CSO - Automated Burning Detection The Village Monitoring System program conducts daily scans of moderate resolution commercial satellite imagery to identify anomalies using the near-infrared band. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (time unknown) No impact 
USDOS CSO - Automated Damage Assessments The Conflict Observatory program analyzes moderate and high-resolution commercial satellite imagery to document a variety of war crimes and other abuses in Ukraine, including automated damage assessments of a variety of buildings, including critical infrastructure, hospitals, schools, crop storage facilities. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (in production less than a year) No impact 
USDVA Physical Therapy App A data source agnostic tool which takes input from a variety of wearable sensors and then analyzes the data to give feedback to the physical therapist in an explainable format. Unclear Monitoring or Detection Unknown Direct impact 
USDVA Coach in Cardiac Surgery Infers misalignment in team members’ mental models during complex healthcare task execution. Of interest are safety-critical domains (e.g., aviation, healthcare), where lack of shared mental models can lead to preventable errors and harm. Identifying model misalignment provides a building block for enabling computer-assisted interventions to improve teamwork and augment human cognition in the operating room. Unclear Organization & Efficiency Unknown Direct impact 
USDVA AI Cure A phone app that monitors adherence to orally prescribed medications during clinical or pharmaceutical sponsor drug studies. Unclear Monitoring or Detection Unknown Direct impact 
USDVA Acute Kidney Injury Focuses on detecting acute kidney injury (AKI), ranging from minor loss of kidney function to complete kidney failure. The artificial intelligence can also detect AKI that may be the result of another illness. Unclear Monitoring or Detection Unknown Direct impact Project is in collaboration with Google DeepMind 
USDVA Assessing lung function in health and disease Determines predictors of normal and abnormal lung function and sleep parameters. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA Automated eye movement analysis and diagnostic prediction of neurological disease Recursively analyzes previously collected data to both improve the quality and accuracy of automated algorithms, as well as to screen for markers of neurological disease (e.g. traumatic brain injury, Parkinson's, stroke, etc). Unclear Monitoring or Detection Unknown Direct impact 
USDVA Automatic speech transcription engines to aid scoring neuropsychological tests. Automated speech transcription engines analyze the cognitive decline of older VA patients. Digitally recorded speech responses are transcribed using multiple artificial intelligence-based speech-to-text engines. The transcriptions are fused together to reduce or obviate the need for manual transcription of patient speech in order to score the neuropsychological tests. Speech-to-Text Organization & Efficiency Unknown Direct impact 
USDVA Curapatient Allows patients to better manage their conditions without having to see a provider. It allows patients to create a profile to track their health, enroll in programs, manage insurance, and schedule appointments. Unclear Organization & Efficiency Unknown Direct impact 
USDVA Digital Command Center Seeks to consolidate all data in a medical center and apply predictive prescriptive analytics to allow leaders to better optimize hospital performance. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA Disentangling dementia patterns using artificial intelligence on brain imaging and electrophysiological data Predict the various patterns of dementia seen on MRI and EEG and explore the use of these imaging modalities as biomarkers for various dementias and epilepsy disorders. The VA is performing retrospective chart review to achieve this. Neural Networks Forecasting & Prediction Unknown Direct impact 
USDVA Enhanced diagnostic error detection and ML classification of protein electrophoresis text Researchers are performing chart review to collect true/false positive annotations and construct a vector embedding of patient records, followed by similarity-based retrieval of unlabeled records "near" the labeled ones (semi-supervised approach). Embedding inputs will be selected high-value structured data pertinent to stroke risk and possibly selected text notes. Support Vector Machines Organization & Efficiency Unknown Indirect impact 
USDVA Behavidence Veterans download the app onto their phone and it compares their phone usage to that of a digital phenotype that represents people with confirmed diagnosis of mental health conditions. Unclear Monitoring or Detection Unknown Direct impact Seems very invasive 
USDVA Tools to predict outcomes of hospitalized VA patients An IRB-approved study which aims to examine machine learning approaches to predict health outcomes of VA patients. It will focus on the prediction of Alzheimer's disease, rehospitalization, and Chlostridioides difficile infection. Machine Learning (Type Unknown) Forecasting & Prediction Short-term project or study Direct impact 
USDVA Nediser A continuously trained “radiology resident” that assists radiologists in confirming the X-ray properties in their radiology reports. Nediser can select normal templates, detect hardware, evaluate patella alignment and leg length and angle discrepancy, and measure Cobb angles. Unclear Monitoring or Detection Unknown Direct impact 
USDVA Precision medicine PTSD and suicidality diagnostic and predictive tool Interprets real time inputs to forewarn episodes of PTSD and suicidality, support early and accurate diagnosis of the same, and gain a better understanding of the short and long term effects of stress, as it relates to the onset of PTSD. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA Prediction of Veterans' Suicidal Ideation following Transition from Military Service Model uses relevant data from a web-based survey of veterans’ experiences within three months of separation and every six months after for the first three years after leaving military service to predict veterans' suicidal ideation. Machine Learning (Type Unknown) Forecasting & Prediction Unknown Direct impact 
USDVA PredictMod Determines if predictions can be made about diabetes based on the gut microbiome. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA Predictor Profiles of OUD and overdose Evaluates the interactions of known and novel risk factors for opioid use disorder (OUD) and overdose in Post-9/11 Veterans. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Unknown Direct impact 
USDVA Provider directory data accuracy and system of record alignment AI is used to add value as a transactor for intelligent identity resolution and linking. AI also has a domain cache function that can be used for both Clinical Decision Support and for intelligent state reconstruction over time and real-time discrepancy detection. As a synchronizer, AI can perform intelligent propagation and semi-automated discrepancy resolution. Unclear Organization & Efficiency Unknown No impact 
USDVA Seizure detection from EEG and video Uses EEG and video data from a VHA epilepsy monitoring unit in order to automatically identify seizures without human intervention. Machine Learning (Type Unknown) Monitoring or Detection Unknown Direct impact 
USDVA SoKat Suicidial Ideation Detection Engine Improves identification of Veteran suicide ideation from survey data collected by the Office of Mental Health Veteran Crisis Line support team. Natural Language Processing Monitoring or Detection Unknown Indirect impact 
USDVA Predict perfusionists’ critical decision-making during cardiac surgery Builds predictive models of perfusionists’ decision-making during critical situations that occur in the cardiopulmonary bypass phase of cardiac surgery. Machine Learning (Type Unknown) Forecasting & Prediction Unknown Direct impact Results may inform future development of computerized clinical decision support tools to be embedded into the operating room, improving patient safety and surgical outcomes. 
USDVA Gait signatures in patients with peripheral artery disease Previously collected biomechanics data is used to identify representative gait signatures of PAD to 1) determine the gait signatures of patients with PAD and 2) the ability of limb acceleration measurements to identify and model the meaningful biomechanics measures from PAD data. Machine Learning (Type Unknown) Monitoring or Detection Unknown Direct impact 
USDVA MedSafe Clinical Decision Support Analyzes current clinical management for diabetes, hypertension, and chronic kidney disease, and makes patient-specific, evidence-based recommendations to primary care providers. Unclear Monitoring or Detection Unknown Direct impact The system uses knowledge bases that encode clinical practice guideline recommendations and an automated execution engine to examine multiple comorbidities, laboratory test results, medications, and history of adverse drug events in evaluating patient clinical status and generating patient-specific recommendations 
USDVA Prediction of health outcomes, including suicide death, opioid overdose, and decompensated outcomes of chronic diseases Using electronic health records (EHR) (both structured and unstructured data) as inputs, this tool outputs deep phenotypes and predictions of health outcomes including suicide death, opioid overdose, and decompensated outcomes of chronic diseases. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA VA-DoE Suicide Exemplar Project Improves VA's ability to identify Veterans at risk for suicide through three closely related projects that all involve collaborations with the Department of Energy. Unclear Monitoring or Detection Unknown Direct impact 
USDVA Disease progression of hepatitis C virus Predicts disease progression among veterans with hepatitis C virus. Machine Learning (Type Unknown) Forecasting & Prediction Unknown Direct impact 
USDVA Biologic response to thiopurines Predictsbiologic response to thiopurines among Veterans with irritable bowel disease. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA Predicting hospitalization and corticosteroid use as a surrogate for IBD flares Examines data from 20,368 Veterans Health Administration patients with an irritable bowel disease diagnosis between 2002 and 2009. Longitudinal labs and associated predictors were used to predict hospitalizations and steroid usage as a surrogate for IBD Flares. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Short-term project or study Direct impact 
USDVA Predicting corticosteroid free endoscopic remission with Vedolizumab in ulcerative colitis Predicts the outcome of corticosteroid-free biologic remission at week 52 on the testing cohort. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Short-term project or study Direct impact This work is focused on a cohort of 594 patients. Models were constructed using baseline data or data through week 6 of VDZ therapy. 
USDVA Predict surgery in Crohn’s disease Analyzes patient demographics, medication use, and longitudinal laboratory values collected between 2001 and 2015 from adult patients in the Veterans Integrated Service Networks 10 cohort. The data was used for analysis in prediction of Crohn’s disease and to model future surgical outcomes within one year. Machine Learning (Type Unknown) Forecasting & Prediction Short-term project or study Direct impact 
USDVA Reinforcement learning evaluation of treatment policies for patients with hepatitis C virus Predicts disease progression among veterans with hepatitis C virus. Machine Learning (Type Unknown) Forecasting & Prediction Unknown Direct impact 
USDVA Predicting hepatocellular carcinoma in patients with hepatitis C Examines whether deep learning recurrent neural network (RNN) models that use raw longitudinal data extracted directly from electronic health records outperform conventional regression models in predicting the risk of developing hepatocellular carcinoma (HCC). Neural Networks Forecasting & Prediction Short-term project or study No impact This prognostic study used data on patients with hepatitis C virus (HCV)-related cirrhosis in the national Veterans Health Administration who had at least 3 years of follow-up after the diagnosis of cirrhosis. 
USDVA Computer-aided detection and classification of colorectal polyps The models receive video frames from colonoscopy video streams and analyze them in real time in order to (1) detect whether a polyp is in the frame and (2) predict the polyp's malignant potential. Unclear Forecasting & Prediction Short-term project or study Direct impact 
USDVA GI Genius (Medtronic) Aids in detection of colon polyps. Unclear Monitoring or Detection Unknown Direct impact 
USDVA Extraction of family medical history from patient records Uses TIU documentation on African American Veterans aged 45-50 to extract family medical history data and identify Veterans who are are at risk of prostate cancer but have not undergone prostate cancer screening. Unclear Forecasting & Prediction Short-term project or study Direct impact 
USDVA VA/IRB approved research study for finding colon polyps Uses a randomized trial for finding colon polyps with artifical intelligence. Unclear Monitoring or Detection Short-term project or study Indirect impact 
USDVA Interpretation/triage of eye images Triages eye patients cared for through telehealth, interprets eye images, and assesses health risks based on retina photos. The goal is to improve diagnosis of a variety of conditions, including glaucoma, macular degeneration, and diabetic retinopathy. Automated Image Processing Monitoring or Detection Unknown Direct impact 
USDVA Screening for esophageal adenocarcinoma National VHA administrative data is used to adapt tools that use electronic health records to predict the risk for esophageal adenocarcinoma. Unclear Forecasting & Prediction Unknown Direct impact 
USDVA Social determinants of health extractor AI is used with clinical notes to identify social determinants of health (SDOH) information. The extracted SDOH variables can be used during associated health related analysis to determine, among other factors, whether SDOH can be a contributor to disease risks or healthcare inequality. Unclear Forecasting & Prediction Unknown Direct impact 
EPA Predict exposure pathways Chemical structure and physicochemical properties were used to predict the probability that a chemical might be associated with any of four exposure pathways leading from sources-consumer (near-field), dietary, far-field industrial, and far-field pesticide-to the general population. We then used exposure pathways to organize predictions from 13 different exposure models as well as other predictors of human intake rates. We created a consensus, meta-model using the Systematic Empirical Evaluation of Models framework. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Short-term project or study Indirect impact The balanced accuracies of these source-based exposure pathway models range from 73 to 81%, with the error rate for identifying positive chemicals ranging from 17 to 36%. 
EPA Records categorization Predict the retention schedule for records; the model will be incorporated into a records management application to help users apply retention schedules when they submit new records. Machine Learning (Type Unknown) Forecasting & Prediction Ongoing project (time unknown) No impact 
EPA Enforcement Targeting Improves enforcement of environmental regulations through facility inspections by the EPA and state partners. Unclear Forecasting & Prediction Ongoing project (time unknown) No impact The resulting predictive analytics showed a 47% improvement of identifying violations of the Resource Conservation and Recovery Act. 
HHS Health Resources and Services Administration Electronic Handbooks AI Chatbot Built to allow grantees to communicate with the EHBs Chatbot using regular natural conversational expressions; provides knowledge- and action-based responses through a self-service platform with 24/7 availability Natural Language Processing Chat Bot Ongoing project (time unknown) Direct impact 
HHS Health Resources and Services Administration BHW Community Need Analysis Platform Allows for BHW to dynamically assess the healthcare need of a population given a specific use case and relevant datasets. The output of the model will be used as part of the Notice of Funding Opportunity (NOFO) grant proposal evaluation process. Machine Learning (Type Unknown) Forecasting & Prediction Ongoing project (time unknown) Indirect impact The first use case being developed is for primary care with behavioral health integration which uses a machine learning based automated clustering engine. 
HHS CDC - ICD-10 Coding of Cause of Death reported on Death Certificates (MedCoder) MedCoder ICD-10 cause of death codes to the literal text cause of death description provided by the cause of death certifier on the death certificate. Unclear Organization & Efficiency Unknown No impact 
HHS CDC - Item Nonresponse Detection in Open-text Response Data Developing an item nonresponse detection model, to identify cases of item nonresponse (e.g., gibberish, uncertain/don’t know, refusals, or high-risk) among open-text responses to help improve survey data and question and questionnaire design. Natural Language Processing Organization & Efficiency Planning or development stage No impact The system is a Natural Language Processing (NLP) model pre-trained using Contrastive Learning and fine-tuned on a custom dataset from survey responses. 
HHS CDC - Sequential Coverage Algorithm (SCA) in Record Linkage Used to develop joining methods (or blocking groups) when working with very large datasets. The SCA method improved the efficiency of blocking. Machine Learning (Type Unknown) Organization & Efficiency Ongoing project (time unknown) No impact 
HHS CMS - Chatbot - Voice Assists the CMS Badging Help Desk with an automated phone response for general badging questions allowing help desk personnel to assist employees and contractors with more detailed/larger issues. Natural Language Processing Chat Bot Ongoing project (time unknown) No impact 
HHS CMS - Chatbot - Text Assists the Security team with an automated email response for general physical security questions, allowing the help desk team to assist employees and contractors with more in depth issues. Natural Language Processing Chat Bot Ongoing project (time unknown) No impact 
HHS CMS - Feedback Analysis Solution Uses CMS or other publicly available data (such as Regulations.Gov) to review public comments and/or analyze other information from internal and external stakeholders. Natural Language Processing Organization & Efficiency Unknown No impact The FAS uses Natural Language Processing (NLP) tools to aggregate, sort and identify duplicates to create efficiencies in the comment review process. FAS also uses machine learning (ML) tools to identify topics, themes and sentiment outputs for the targeted dataset. 
HHS CMS - Predictive Intelligence - Incident Assignment for Quality Service Center Analyzes the short description provided by the end user in order to find key words with previously submitted incidents and assigns the ticket to the appropriate assignment group. Natural Language Processing Organization & Efficiency Unknown Indirect impact Predictive Intelligence (PI) is used for incident assignment within the Quality Service Center (QSC). The solution runs on incidents created from the ServiceNow Service Portal. This solution is re-trained with the incident data in our production instance every 3-6 months based on need. 
HHS CMS - Reasonable Accomodation RPA Bot Pulls HR data related to staffing changes, e.g. promotions, reassignments, change in supervisor, and generates information for action by Reasonable Accommodation staff to ensure disability reasonable accommodations follow the employee. Natural Language Processing Organization & Efficiency Unknown No impact 
HHS CMS - Rapid Authority to Operate Used to identify common blocks of language used in similar ways across system security plan (SSP) documents. CMS could identify similar approaches to solving certain technology or process-related control areas within the Acceptable Risk Safeguards. The output was used to create a list of components to develop control description language in a re-usable way, as part of the Blueprint/Rapid ATO effort to streamline SSP generation for new systems. Natural Language Processing Security Unknown No impact 
HHS CMS - Data Lake/Load-Extract-Load-Transform (L-ETL) Modernizes the load-extract-load-transform (L-ETL) pipelines and data tooling. CMS will be enhancing Agency security to bring together more system, telemetry and program data in one place with a unifying governance model. High-Power Computing Security Planning or development stage No impact There is no actual ML/AI work being done here today, rather, we are beginning work on the scaffolding that will open up these opportunities in 1-2 years time. 
HHS CMS - Priority Score Model Ranks providers within the Fraud Prevention System using logistic regression based on program integrity guidelines. Regression Analysis Mapping Unknown Indirect impact Inputs - Medicare Claims data, Targeted Probe and Educate (TPE) Data, Jurisdiction information 
HHS CMS - Priority Score Timeliness Forecast the time needed to work on an alert produced by Fraud Prevention System Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Unknown No impact 
HHS CMS - Provider Education 90 Day Reviews claims for provider before and after education for statistical change in their claim submission patterns Regression Analysis Research (Other) Unknown No impact 
HHS FDA - Advanced Semantic Search and Indexing of Text for Tobacco Applications (ASSIST4Tobacco) Uses semantic indexing to search tobacco authorization applications. Natural Language Processing Organization & Efficiency Planning or development stage No impact 
HHS FDA - Artificial Intelligence-based Deduplication Algoirthm for Classfication of Duplicate Reports in the FDA Adverse Event Reports (FAERS) The deduplication algorithm is applied to nonpublic data in the FDA Adverse Event Reporting System (FAERS) to identify duplicate reports; structured and unstructured data are used in a probabilistic record linkage approach to score pairs of reports by evaluating multiple data fields and applying relative weights per field. Natural Language Processing Organization & Efficiency Unknown No impact The output of potential duplicate reports is further placed in groups to facilitate identification of FAERS reports during case series evaluation for safety issues of concern. 
HHS FDA - Opioid Data Warehouse Term Identification and Novel Synthetic Opioid Detection and Evaluation Analytics Uses publicly available social media and forensic chemistry data to identify novel referents to drug products in social media text. Network Analysis (ie Bayesian, or Social Network) Research (Other) Unknown No impact It uses the FastText library to create vector models of each known NSO-related term in a large social media corpus, and provides users with similarity scores and expected prevalence estimates for lists of terms that could be used to enhance future data gathering efforts 
HHS NIH - National Institute of General Medical Sciences (NIGMS) AI Supported Searches, Information Systems and Tools System Provides the ability to identify investigators by PPID from Federal RePORTER based on user input of investigator PPIDs; provides the ability to lookup potential matching program officers, including their corresponding predicted Program Area Codes, and ICs based on the input of unstructured scientific data. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact DIMA and IRMB have collaborated to develop functions that utilize artificial intelligence and natural language processing methods to produce data relevant to the NIGMS program staff’s mission. These tools are collected into a single system to make them available to the NIGMS community for use on a day-to-day basis. 
HHS NIH - Leveraging AI for Business Process Automation Automates the initial referral of grant applications to the proper scientific expertise within the Institute. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact NIGMS IRMB and DIMA are currently using this NLP/ML algorithm developed in R statistical software to parse grant applications and to determine Project Officer candidates for grant assignment. This process was previously fully manual and required a substantial person hour effort. 
HHS NIH - Grant Application Subject-Matter Classification Tool Classifies grant applications for review assignment. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
HHS NIH - Splunk IT System Monitoring Software Aggregates system logs from IT infrastructure systems and endpoints for auditing and monitoring purposes. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (time unknown) No impact 
HHS NIH - COVID-19 Pandemic Vulnerability Index Dashboard Creates risk profiles, called PVI scorecards, for every county in the United States, continuously updated with the latest data that summarize and visualize overall disease risk. Unclear Mapping Unknown No impact 
HHS NIH - Leveraging AI/ML for classification and categorization of scientific concepts Used for topical characterization of the research portfolio. Machine Learning (Type Unknown) Classification or Labeling Unknown No impact Inputs are publications and grants abstracts. These are fed into a text classification model and concept extraction. The outputs are category labels and list of concepts. 
HHS NIH - Machine learning system to predict translational progress in biomedical research Detects whether a paper is likely to be cited by a future clinical trial or guideline. Machine Learning (Type Unknown) Forecasting & Prediction Unknown No impact Despite the noisiness of citation dynamics, as little as 2 years of postpublication data yield accurate predictions about a paper’s eventual citation by a clinical article (accuracy = 84%, F1 score = 0.56; compared to 19% accuracy by chance). 
HHS NIH - Semantic analysis of scientific documents Computationally converts words in scientific texts to numbers and summarizes documents by their semantic content by learning relationships between words from their context. Neural Networks Research (Other) Unknown No impact This method is adaptable to specific corpora, including grants and scientific articles. 
HHS NIH - Person-level disambiguation for PubMed authors and NIH grant applicants Determines whether author-publication pairs refer to variant representations of the same person; for example, model can determine whether hypothetical records listing Jane Smith and Jane M. Smith were the same person, or two different people, based on variables that include institutional affiliation, co-authorship, and article-affiliated Medical Subject Heading (MeSH) terms. Neural Networks Research (Other) Ongoing project (time unknown) No impact High-quality disambiguation is required to correctly link researchers to their grants and outputs including articles, patents, and clinical trials. 
HHS NIH - Program Class Code (Area of Science) Referral for NIAID Evaluates the projects that are in Referral, Program Analysis Branch and auto assigns these grant applications to the Program Class Codes. Unclear Classification or Labeling Planning or development stage No impact The inputs are comprised of approximately 6,000+ grant applications that are currently manually assigned by RPAB Staff. The output would be grant applications that are categorized into their respective PCC's. 
HHS NIH - Research, Condition, and Disease Categorization RCDC is an electronic budget reporting tool that categorizes projects using AI/NLP. The inputs are grant applications, R&D contracts, intramural projects, inter agency agreements. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
HHS NIH - Query View Report (QVR) LIKE The LIKE feature in QVR makes use of the NIH Research, Condition and Disease Categorization (RCDC) indexing results to compare scientific terms associated with a project, person or publication and find scientifically similar projects, persons or publications. Unclear Classification or Labeling Unknown No impact 
HHS NIH - Internal Referral Module Automatically refers projects to Program Officers once the grant application is received. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact This process, is operating at a high accuracy rate and has effectively eliminated the referral bottleneck. 
HHS NIH Grants Virtual Assistant Chat Bot to assist users in finding grant related information via OER resources. Natural Language Processing Chat Bot Ongoing project (time unknown) No impact 
HHS NIH - Pangolin lineage classifications to support accessing and analysis of SARS-CoV-2 sequence data The Pango nomenclature, called Pango lineages, is being used by researchers and public health agencies worldwide to track the transmission and spread of SARS-CoV-2, including variants of concern. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Monitoring or Detection Unknown No impact 
HHS NIH - Providing MeSH Check Tag of NLM’s Medical Text Indexer (MTI) ons using Support Vector Machines Provides confidence scores for a set of MeSH CheckTags to the NLM Medical Text Indexer (MTI) program; these CheckTags are small set of MeSH Descriptors designed to indicate Species, Sex, and Age in MEDLINE articles. Machine Learning (Type Unknown) Classification or Labeling Unknown No impact 
HHS NIH - Determining selection for indexing MEDLINE articles using Neural Network Architecture with a Convolutional Neural Network Uses a sigmoid activation function to generate a single output value between zero and one, which can be interpreted as the probability of an article being in-scope for MEDLINE. Neural Networks Forecasting & Prediction Unknown No impact 
HHS NIH - MetaMap to identity potential terms for indexing MEDLINE articles Provides a link between the text of biomedical literature and the knowledge, including synonymy relationships, embedded in the Metathesaurus Natural Language Processing Mapping Unknown No impact 
HHS NIH - Best Match: New relevance search for PubMed Best Match is a new relevance search algorithm for PubMed that leverages the intelligence of our users and cutting-edge machine-learning technology as an alternative to the traditional date sort order. Machine Learning (Type Unknown) Research (Other) Ongoing project (time unknown) No impact PubMed is a free search engine for biomedical literature accessed by millions of users from around the world each day. With the rapid growth of biomedical literature, finding and retrieving the most relevant papers for a given query is increasingly challenging. 
HHS NIH - SingleCite: Improving single citation search in PubMed SingleCite is an automated algorithm that establishes a query-document mapping by building a regression function to predict the probability of a retrieved document being the target based on three variables: the score of the highest scoring retrieved document, the difference in score between the two top retrieved documents, and the fraction of a query matched by the candidate citation. Regression Analysis Mapping Ongoing project (time unknown) No impact 
HHS NIH - Computed Author: author name disambiguation for PubMed Author name ambiguity may lead to irrelevant retrieval results; we developed a machine-learning method to score the features for disambiguating a pair of papers with ambiguous names. Machine Learning (Type Unknown) Research (Other) Ongoing project (time unknown) No impact 
HHS NIH - National Library of Medicine NLM-Gene: towards automatic gene indexing in PubMed articles An automatic tool for finding gene names in the biomedical literature. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact 
HHS NIH - National Library of Medicine NLM-Chem: towards automatic chemical indexing in PubMed articles An automatic tool for finding chemical names in the biomedical literature. Natural Language Processing Organization & Efficiency Ongoing project (time unknown) No impact 
HHS OIG - Grants Analytics Portal Enhances staff’s ability to access grants related data quickly and easily by: quickly navigating directly to the text of relevant findings across thousands of audits, the ability to discover similar findings, analyze trends, compare data between OPDIVs, and the means to see preliminary assessments of potential anomalies between grantees. Unclear Organization & Efficiency Ongoing project (time unknown) No impact 
HHS OIG - Text Analytics Portal Allows personnel without an analytics background to quickly examine text documents through a related set of search, topic modeling and entity recognition technologies. Unclear Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Relevancy Tailoring Adjusts the ranking of search results so that most relevant results show up at the top of the list. Unclear Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Auto-generation Synonyms Adds synonyms to search queries to improve search results. Unclear Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Automated Suggestions Auto-fills queries as they are typed. Machine Learning (Type Unknown) Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Suggested Related Content Shows related searches that may provide the user with other related, valuable information. Machine Learning (Type Unknown) Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Auto Tagging Suggests content tags automatically based on a machine-driven evaluation of how existing content is tagged. Machine Learning (Type Unknown) Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Did you mean Suggests spelling corrections and reformatted search queries based on Google Analytics data. Machine Learning (Type Unknown) Organization & Efficiency Ongoing project (time unknown) No impact 
HHS AHRQ - Chatbot Responds to plain language queries in real time. Natural Language Processing Research (Other) Ongoing project (time unknown) No impact 
DOJ Drug Signature Program Algorithms Automatically classifies the geographical region of origin of samples selected for DEA's Heroin and Cocaine signature programs. The system provides for detection of anomalies and low confidence results. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production more than a year) No impact Agency-generated Data not available publicly but 1-2 unclassified summary reports for each program are released publicly each year. 
DOJ Complaint Lead Value Probability Helps to triage immediate threats in order to help FBI field offices and law enforcement respond to the most serious threats first. Based on the algorithm score, highest priority tips are first in the queue for human review. Unclear Classification or Labeling Ongoing project (in production more than a year) Indirect impact Agency-generated Threat Intake Processing System (TIPS) database uses artificial intelligence (AI) algorithms to accurately identify, prioritize, and process actionable tips in a timely manner. 
DOJ Intelligent Records Consolidation Tool Assesses the similarity of records schedules across all Department records schedules. The tool provides clusters of similar items to significantly reduce the time that the Records Manager spends manually reviewing schedules for possible consolidation. Natural Language Processing Organization & Efficiency Ongoing project (in production more than a year) No impact An AI powered dashboard provides recommendations for schedule consolidation and review, while also providing the Records Manager with the ability to review by cluster or by individual record. 
DOJ Privileged Material Identification Scans documents and looks for attorney/client privileged information. It does this based on keyword input by the system operator. Optical Character Recognition (or Text Extraction) Monitoring or Detection Ongoing project (in production less than six months) No impact 
DOT Technical Operations Predictive Maintenance Utilize equipment telemetry data and statistical modeling to predict equipment failures before they occur in order to improve operational efficiency and safety by reducing unscheduled outages and/or shortening outage times as replacement equipment can be pre-positioned in anticipation of the failure. Machine Learning (Type Unknown) Organization & Efficiency Planning or development stage Indirect impact 
DOT Surface Report Classifiier (SCM/Auto-Class) SCM classifies surface incident reports by event type, such as Runway Incursion, Runway Excursion, Taxiway Incursion/Excursion and categorizes runway incursions further by severity type. Support Vector Machines Classification or Labeling Planning or development stage No impact 
DOT Regulatory Compliance Mapping Tool RCMT processes all the documents’ paragraphs to extract the meaning (semantics) of the text. RCMT then employs a recommender system (also using some AI technology) to take the texts augmented by the texts’ meaning to establish candidate matches between the ICAO SARPs and FAA text that provides means of compliance. Natural Language Processing Organization & Efficiency Ongoing project (in production less than a year) No impact The AVS International office is required to identify means of compliance to ICAO Standards and Recommended Practices (SARPs). Both SARPs and means of compliance evidence are text paragraphs scattered across thousands of pages of documents. AOV identified a need to find each SARP, evaluate the text of many FAA Orders, and suggest evidence of compliance based upon the evaluation of the text. The base dataset used by RCMT is the documents’ texts deconstructed into paragraphs. 
DOT Fusion and analysis of safety event reporting data Integrates safety event reporting data from aircraft operators and manufacturers in support of data-driven decision making. The development of an ontology helped standardize and integrate data from across disconnected sources. Natural Language Processing Organization & Efficiency Planning or development stage No impact 
DOT JASC Code classification in Safety Difficulty Reports (SDR) Derives the joint aircraft system codes (JASC) chapter codes from the narrative description within service difficulty reports (SDR), a form of safety event reporting from aircraft operators. Natural Language Processing Classification or Labeling Planning or development stage No impact 
DOT Safety risk classification Uses features from Safety Management Tracking System (SMTS) hazards, projects, Safety Risk Management Documents (SRMDs), requirements, and SPT's to predict an initial risk level. Natural Language Processing Forecasting & Prediction Planning or development stage No impact 
DOT Feature importance modeling Uses regression models to determine statistical correlations between AOV indicators with AJI's ASM and Surface Safety Metric (SSM) metrics. Machine Learning (Type Unknown) Research (Other) Planning or development stage No impact 
DOT Anomaly Detection Detects variances in Remote Monitoring and Logging System (RMLS) log outages to predict whether an outage should or should not be an SIE. Neural Networks Forecasting & Prediction Planning or development stage No impact 
DOT Use Case to Identify Energy Signatures Leveraging fusion data to correlate unstable approaches to energy “signatures” on approach. Unclear Research (Other) Planning or development stage No impact 
DOT Offshore Precipitation Capability (OPC) OPC leverages data from several sources such as weather radar, lightning networks, satellite and numerical models to produce a radar-like depiction of precipitation. The algorithm then applies machine learning techniques based on years of satellite and model data to improve the accuracy of the location and intensity of the precipitation areas. Machine Learning (Type Unknown) Mapping Ongoing project (in production more than a year) No impact 
DOT Course Deviation Identification for Multiple Airport Route Separation (MARS) May enable deconfliction of airports in high-demand metropolitan areas. To build necessary collision risk models for the safety case, several models are needed, including one that describes the behavior of aircraft that fail to navigate the procedure correctly. Machine Learning (Type Unknown) Forecasting & Prediction Planning or development stage Indirect impact 
DOT Determining Surface Winds with Machine Learning Software Analyzes camera images of a wind sock to produce highly accurate surface wind speed and direction information in remote areas that don’t have a weather observing sensor. Automated Image Processing Mapping Planning or development stage No impact 
DOT Remote Oceanic Meteorological Information Operations (ROMIO) Evaluates the feasibility to uplink convective weather information to aircraft operating over the ocean and remote regions. Capability converted weather satellite data, lightning and weather prediction model data into areas of thunderstorm activity and cloud top heights. Neural Networks Mapping Planning or development stage No impact 
SSA Insight Analyzes the free text of disability decisions and other case data to offer adjudicators real-time alerts on potential quality issues and case-specific reference information within a web application; offers adjudicators a series of interactive tools to help streamline their work. Natural Language Processing Monitoring or Detection Unknown Indirect impact 
SSA Intelligent Medical Langage Analysis Generation (IMAGEN) Analyzes clinical text from disability applicants health records and transforms it to data and other useful formats to enable disability adjudicators to more easily find and identify clinical content that is relevant to SSA’s disability determination process. Natural Language Processing Classification or Labeling Unknown Indirect impact 
SSA Duplicate Identification Process (DIP) Helps the user to identify and flag duplicate pages and documents within the disability electronic folder more efficiently, reducing the amount of task time associated with preparing cases for SSA's ALJ Hearings. Automated Image Processing Organization & Efficiency Unknown No impact 
SSA Handwriting recognition from forms Parses handwritten entries on specific standard forms submitted by clients. Optical Character Recognition (or Text Extraction) Organization & Efficiency Unknown Indirect impact 
GSA Acquisition Analytics Takes Detailed Data on transactions and classifies each transaction within the Government-wide Category Management Taxonomy. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
GSA Category Taxonomy Refinement Uses token extraction from product descriptions more accurately shape intended markets for PSCs. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
GSA Chatbot for Federal Acquisition Community Streamlines the customer experience process, and automates providing answers to documented commonly asked questions through public facing knowledge articles. Natural Language Processing Chat Bot Planning or development stage Direct impact 
GSA City Pairs Program Ticket Forecast and Scenario Analysis Tools Takes segment-level City Pair Program air travel purchase data and creates near-term forecasts for the current and upcoming fiscal year by month and at various levels of granularity including DOD vs Civilian, Agency, and Region. Unclear Forecasting & Prediction Planning or development stage No impact 
GSA Classifying Qualitative Data with Medallia Users can create rules based on words and their relationships with other words to tag qualtitative data with our topics (passports, tax refunds, etc.); also offers sentiment analysis. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
GSA Contract Acquisition Lifecycle Intelligence (CALI) Streamlines the evaluation of vendor proposals. Machine Learning (Type Unknown) Project Management Planning or development stage No impact Offered by Octo Consulting 
GSA Enterprise Brain Document repository that improves document discovery. Unclear Monitoring or Detection Planning or development stage No impact Document repository from tanjo.ai 
GSA Key KPI Forecasts for GWCM Takes monthly historical data for underlying components used to calculate KPIs and creates near-term forecasts for the upcoming fiscal year. Unclear Forecasting & Prediction Planning or development stage No impact 
GSA OAS Kudos Chatbot Captures employee peer-to-peer recognition. Natural Language Processing Chat Bot Planning or development stage No impact 
GSA ServiceNow Generic Ticket Classification Takes generic Service Now tickets and classify them so that they can be automatically re-routed to the correct team that handles these types of tickets. Natural Language Processing Classification or Labeling Planning or development stage No impact 
GSA Solicitation Review Tool (SRT) The SRT intakes SAM.gov data for all ICT solicitations. The system then compiles the data into a database to be used by machine learning algorithms. The first of these is a Natural Language Processing model that determines if a solicitation contains compliance language. If a solicitation does not have compliance language, then it is marked as non-compliant. Natural Language Processing Classification or Labeling Ongoing project (time unknown) No impact 
GSA Survey Comment Ham / Spam tester Determines which comments on USA.gov comments section are worth the time of analysts reading. Natural Language Processing Classification or Labeling Planning or development stage Indirect impact 
DHS Sentiment Analysis and Topic Modeling (SenTop) Intially, analyzed survey responses for DHS’s Office of the Chief Procurement Officer related to contracting; currently, serves as general-purpose text analytics solution that can be applied to any domain/area. Natural Language Processing Project Management Ongoing project (in production more than a year) No impact 
DHS CISA - AIS Scoring & Feedback (AS&F) AIS enables the real-time exchange of machine-readable cyber threat indicators and defensive measures to help protect against and ultimately reduce the prevalence of cyber incidents; AS&F specifically performs descriptive analytics from organizational-centric intelligence to support confidence and opinion classification of indicators of compromise. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production less than six months) No impact 
DHS CISA - Automated PII Detection Automatically detects potential PII from within Automated Indicator Sharing submissions. If submissions are flagged for possible PII, the submission will be queued for human review where the analysts will be provided with the submission and artificial intelligence-assisted guidance to the specific PII concerns Natural Language Processing Monitoring or Detection Ongoing project (in production less than six months) Indirect impact 
DHS TSA - CDC Airport Hotspot Throughput (PageRank) Determines the domestic airports that have the highest rank of connecting flights during the holiday travel season to help mitigate the spread of COVID-19. Unclear Mapping Ongoing project (in production more than a year) No impact This capability is a DHS-developed artificial intelligence model written in Spark/Scala that takes historical non-PII travel data and computes the highest-ranking airports based on the PageRank algorithm. 
DHS USCG - Silicon Valley Innovation Program (SVIP) Language Translator Supports the Coast Guard in facilitating real-time communications with non-English speakers and those who are unable to communicate verbally. The solicitation also included requirements for language translation technology to be capable of operating both online and offline because many Coast Guard interactions take place in extreme environmental conditions, and in locations without cell service or an internet connection. Speech-to-Text Organization & Efficiency Ongoing project (in production more than a year) No impact 
DHS USCIS - Asylum Text Analytics (ATA) Identifies plagiarism-based fraud in applications for asylum status and for the withholding of removal by scanning the digitized narrative sections of the associated forms and looking for common language patterns. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (in production more than a year) Indirect impact 
DHS USCIS - BET/FBI Fingerprint Success Maximization Enables technicians to receive immediate feedback when a set of prints is likely to be rejected by the FBI; aims to maximize the number of successful FBI submissions while minimizing the number of fingerprint recaptures necessary. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production less than a year) Indirect impact USCIS's Customer Profile Management Service (CPMS) serves as a person-centric repository of biometric and biographic information provided by applicants and petitioners (hereafter collectively referred to as “benefit requestors”) that have been issued a USCIS card evidencing the granting of an immigration related benefit (i.e., permanent residency, work authorization, or travel documents). 
DHS USCIS - Biometrics Enrollment Tool (BET) Fingerprint Quality Score Takes a fingerprint image and assigns a score between 0 - 100, with 100 indicating that this is the best quality fingerprint image that could be obtained. The higher the score, the more likely that the fingerprint will match when captured again. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (time unknown) Indirect impact 
DHS USCIS - Evidence Classifier Systematically tags individual pages with some of the highest-volume, highest-impact evidence types to help case workers sort through certain immigration request forms. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production more than a year) No impact 
DHS USCIS - FDNS-DS NexGen Aids the Fraud Detection and National Security (FDNS) Directorate in investigative work, enhances investigative case prioritization, and detects duplicate case work. Machine Learning (Type Unknown) Research (Other) Ongoing project (in production more than a year) Indirect impact 
DHS USCIS - Sentiment Analysis USCIS issued a two-part survey asking users both quantitative and qualitative questions and then assigned "sentiments" to categories ranging from strongly positive to strongly negative. Natural Language Processing Monitoring or Detection Ongoing project (in production more than a year) Indirect impact 
DHS USCIS - Testing Performance of ML Model using H2O Determines the most used categories for applicants submitting I-90's, and machine learning to create predictions of workloads. Machine Learning (Type Unknown) Forecasting & Prediction Ongoing project (in production more than a year) Indirect impact 
DHS USCIS - Timeseries Analysis and Forecasting Used Autoregressive Integrated Moving Average (ARIMA) models on the I-90 form, which allowed the prediction of the total number of forms for a 2-year period. Regression Analysis Forecasting & Prediction Ongoing project (in production more than a year) No impact 
DHS CBP - Agent Portable Surveillance Identifies border activities of interest by analyzing data from Electro-Optical/Infra-Red cameras and radar. Machine Learning (Type Unknown) Security Ongoing project (in production more than a year) Indirect impact 
DHS CBP - Autonomous Surveillance Towers Scans constantly and autonomously; radar detects and recognizes movement; and the camera slews autonomously to the items of interest and the system software identifies the object; analyzes the camera and radar data which alerts the user and autonomously tracks the item of interest. Machine Learning (Type Unknown) Monitoring or Detection Ongoing project (in production more than a year) Indirect impact 
DHS CBP - I4 Viewer Matroid Image Analysis Enables CBP end users to create and share vision detectors; Matroid detectors are trained computer vision models that recognize objects, people, and events in any image and in video streams. Automated Image Processing Monitoring or Detection Ongoing project (time unknown) Indirect impact 
DHS CBP - Open-source News Aggregation Enables users to make better decisions faster by identifying and forecasting emerging events on a global scale to mitigate risk, recognize threats, greatly enhance indications and warnings, and provide predictive intelligence capabilities Network Analysis (ie Bayesian, or Social Network) Forecasting & Prediction Ongoing project (in production more than a year) No impact 
DHS ICE - Data Tagging and Classification RAVEn leverages data tracking and classification to do the following: streamline how special agents and criminal analysts search, filter, translate, and report on electronic communications evidence and will help investigators more effectively determine the structure and organization of criminal enterprises; send and receive leads and enter outcomes such as arrests and seizures; improve the efficiency of agents and analysts in identifying pertinent evidence, relationships, and criminal networks from data extracted from mobile devices. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production less than six months) No impact The Homeland Security Investigations (HSI) Innovation Lab is developing an analytical platform called the Repository for Analytics in a Virtualized Environment (RAVEn). RAVEn facilitates large, complex analytical projects to support ICE’s mission to enforce and investigate violations of U.S. criminal, civil, and administrative laws. 
DHS ICE - Language Translator Used to increase the efficiency, accuracy, and quality of searching, analyzing, and translating speech. Natural Language Processing Classification or Labeling Ongoing project (in production more than a year) No impact 
DHS ICE - RAVEn Compliance Automation Tool (CAT) Increase the speed and efficiency of ingesting and processing Forms I-9 data. Optical Character Recognition (or Text Extraction) Organization & Efficiency Ongoing project (in production less than six months) Indirect impact RAVEn CAT currently employs an Optical Recognition Service (OCR) model and software (Tesseract OCR) to identify pixel coordinates of handwritten and read/extract computer typed characters from ingested forms for processing. 
DHS ICE - RAVEn Normalization Services Service to verify, validate, correct, and normalize the accuracy and quality of addresses, phone numbers, and names. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (in production more than a year) Indirect impact The normalization services let agents analyze both well-defined addresses (such as those in CONUS and Europe) and less well-defined addresses (such as addresses using mile markers); standardize phone numbers to their identified country and to the E164 ITU standard; and streamline the process of correcting data entry errors and/or pointing out purposeful misidentification, connecting information about a person across HSI datasets, and cutting down the number of resource hours needed for investigations. 
DOE Advances in Nuclear Fuel Cycle Nonproliferation, Safeguards, and Security Using an Integrated Data Science Approach Develops a digital twin of a centrifugal contactor system that receives data from traditional and real time sensors, constructs a digital representation or simulation of the chemical separations component within the nuclear fuel cycle, and performs data analysis through machine learning to determine anomalies, failures, and trends. Machine Learning (Type Unknown) Project Management Planning or development stage No impact 
DOE Development of a multi-sensor data science system used for signature development on solvent extraction processes conducted within Beartooth facility Utilizes non-traditional measurement sources such as vibration, acoustics, current, and light, and traditional sources such as flow, and temperature in conjunction with data-based, machine learning techniques that will allow for signal discovery. The goal is to characterize stages within a solvent extraction process can increase target metals recovery, indicate process faults, account for special nuclear material, and inform near real-time decision making. Machine Learning (Type Unknown) Project Management Planning or development stage No impact 
DOE Scalable Framework of Hybrid Modeling with Anticipatory Control Strategy for Autonomous Operation of Modular and Microreactors Validates novel and scalable models to achieve faster-than-real-time prediction and decision-making capabilities; analyzes the risk of cascading failures when emerging reactors are deployed as part of a full feeder microgrid. Machine Learning (Type Unknown) Forecasting & Prediction Unknown No impact 
DOE Accelerating and Improving the Reliability of Low Failure Probability Computations to Support the Efficient Safety Evaluation and Deployment of Advanced Reactor Technologies Reduces the computational burden by reducing the number of finite element evaluations when estimating low failure probabilities. Unclear Organization & Efficiency Planning or development stage No impact These will be implemented in the Multiphysics Object-Oriented Simulation Environment, which will help the nuclear engineering community to efficiently conduct probabilistic failure analyses and uncertainty quantification studies for the design and optimization of advanced reactor technologies. 
DOE Accelerating deployment of nuclear fuels through reduced-order thermophysical property models and machine learning Develops a novel physics-based tool that combines 1) reduced-order models, 2) machine learning algorithms, 3) fuel performance methods, and 4) state-ofthe-art thermal property characterization equipment and irradiated nuclear fuel data sets to accelerate nuclear fuel discovery, development, and deployment. Machine Learning (Type Unknown) Research (Other) Planning or development stage No impact 
DOE Promoting Optimal Sparse Sensing and Sparse Learning for Nuclear Digital Twins Addresses the efficient use of limited experimental data available for nuclear digital twin (NDT) training and demonstration. This involves developing sparse data reconstruction methods and using NDT models to define sensor requirements (location, number, accuracy) for the design of demonstration experiments. Unclear Research (Other) Planning or development stage No impact 
DOE Artificial Intelligence Enhanced Advanced Post Irradiation Examination Uncovers the relationships between micro/nanoscale structure, zirconium phase redistribution, local thermal conductivity, and engineering scale fuel properties; shows how artificial intelligence (AI)-based technology can facilitate and accelerate nuclear fuel development. Neural Networks Research (Other) Unknown No impact 
DOE Secure Millimeter Wave Spectrum Sharing with Autonomous Beam Scheduling Exploits the millimeter wave beam directionality and utilizes the beam sensing capabilities at end devices to prove that an autonomous radio frequency beam scheduler can support secure 5G spectrum sharing and guarantee optimality for base stations. Unclear Research (Other) Unknown No impact 
DOE Objective-Driven Data Reduction for Scientific Workflows Aims to develop theories and algorithms for objective-driven reduction of scientific data in workflows that are composed of various models, including datadriven AI models. Unclear Research (Other) Planning or development stage No impact 
DOE The Grid Resilience and Intelligence Platform (GRIP) Develops metrics that quantify the impact of the anticipated weather related extreme events. The platform uses utility data combined with physical models, distribution power solver to infer the potential grid impacts given a major storm. Unclear Forecasting & Prediction Unknown No impact 
DOE Open-Source High-Fidelity Aggregate Composite Load Models of Emerging Load Behaviors for Large-Scale Analysis (GMLC 0064) Estimates the load composition data and motor protection profiles for different climante regions in the Western US; calibrates the parameters of WECC composite load model to match the responses with detailed feeder model. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Unknown No impact 
DOE Big Data Synchrophasor Monitoring and Analytics for Resiliency Tracking (BDSMART) Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE Combinatorial Evaluation of Physical Feature Engineering and Deep Temporal Modeling for Synchrophasor Data at Scale Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE MindSynchro Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE PMU-Based Data Analytics Using Digital Twin Phasor Analytics Software Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE A Robust Event Diagnostic Platform: Integrating Tensor Analytics and Machine Learning Into Real-time Grid Monitoring Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE Discovery of Signatures, Anomalies, and Precursors in Synchrophasor Data with Matrix Profile and Deep Recurrent Neural Networks Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE Machine Learning Guided Operational Intelligence Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE Robust Learning of Dynamic Interactions for Enhancing Power System Resilience Explore the use of big data tools on phasor measurement unit data to identify and improve existing knowledge, and to discover new insights and tools for better grid operation and management. Machine Learning (Type Unknown) Monitoring or Detection Unknown No impact 
DOE Artificial Intelligence Based Process Control and Optimization for Advanced Manufacturing Develops the capability to intelligently control and optimize advanced manufacturing processes instead of the existing trial and error approach. Neural Networks Project Management Unknown No impact Artificial intelligence-based control algorithms will be developed by employing deep reinforcement learning. To reduce the computational expense with advanced manufacturing models, physics-informed reduced order models (ROMs) will be developed. The AI-based control algorithms will employ the ROMs’ predictions to adaptively inform processing decisions in a simulation environment. 
DOE Smart Contingency Analysis Neural Network for in-depth Power Grid Vulnerability Analyses Machine learning framework and resilience-chaos plots are leveraged to reduce computational expense required to discover, with 90% accuracy, n-2 contingencies by 50%. Machine Learning (Type Unknown) Organization & Efficiency Unknown No impact 
DOE Resilient Attack Interceptor for Intelligent Devices Focuses on developing external monitoring methods to protect industrial internet of things devices by correlating observable physical aspects that are produced naturally and involuntarily during the operational lifecycle with anomalous functionality. Unclear Monitoring or Detection Ongoing project (time unknown) No impact 
DOE Infrastructure eXpression Translates industrial control system features to a machine-readable format for use with automated cyber tools. Unclear Research (Other) Short-term project or study No impact This project’s success can serve as the foundation for prioritizing the next research steps to realize automated threat response, improving the timeliness and fidelity of cyber incident consequence models, and enriching national capabilities to share actionable threat intelligence at machine speed. 
DOE Protocol Analytics to enable Forensics of Industrial Control Systems Discovers methods and technologies to bridge gaps between the various industrial control systems (ICS) communication protocols and standard Ethernet to enable existing cybersecurity tools defend ICS networks and empower cybersecurity analysts to detect compromise before threat actors can disrupt infrastructure, damage property, and inflict harm. Machine Learning (Type Unknown) Research (Other) Unknown Indirect impact 
DOE Automated Type and Data Structure Resolution Identified and labeled type and structure data in an automated and scalable way such that the information can be used in other tools and other Reverse Engineering at Scale research areas such as symbolic execution. Machine Learning (Type Unknown) Classification or Labeling Unknown No impact 
DOE Signal Decomposition for Intrusion Detection in Reliability Assessment in Cyber Resilience Provides a straightforward framework wherein an anomaly detection algorithm can be trained on existing expected data and then used for false data injection detection. Machine Learning (Type Unknown) Monitoring or Detection Planning or development stage No impact An advanced library for signal decomposition and analysis will be developed that allows combining machine learning and artificial intelligence algorithms and high-fidelity model comparisons for greatly improved false data injection detection. This library will facilitate online and posteriori analysis of digital signals for the purpose of detecting potential malicious tampering in physical processes. 
DOE Advanced Machine Learning-based Fifth Generation Network Attack Detection System Proves that enhancing attack detection via innovative machine learning techniques into the fifth generation (5G) cellular network can help to secure mission-critical applications, such as automated vehicles and drones, connected health, emergency response operations, and other missioncritical devices. Machine Learning (Type Unknown) Security Planning or development stage Indirect impact 
DOE Red Teaming Artificial Intelligence Provides methods for the reverse engineering, exploitation, risk assessment and vulnerability remediation. The insights gained from the explorations into vulnerability assessment research will proactively address critical gaps in the cybersecurity community’s understanding of these systems. Machine Learning (Type Unknown) Security Planning or development stage Indirect impact 
DOE Unattended Operation through Digital Twin Innovations Predicts events using the integrated data from test bed sensors and physics-based models; produces a framework for future digital twins. Unclear Forecasting & Prediction Unknown No impact 
DOE Secure and Resilient Machine Learning System for Detecting Fifth Generation (5G) Attacks including Zero-Day Attacks Implements an advanced machine learning based 5G attack detection system that can achieve high classification speed with high accuracy (90% or greater) as well as address a vulnerability to zero-day attacks using field programmable gate array based deep autoencoders. Machine Learning (Type Unknown) Classification or Labeling Unknown No impact 90% accuracy against real zero-day attacks recorded by Amazon Web Services. 
DOE Automated Malware Analysis Via Dynamic Sandboxes Allows for automated analysis, provides non-existing core capabilities to analyze industrial control system malware, and outputs to a format that is machine readable and an industry standard in sharing threat information. Machine Learning (Type Unknown) Research (Other) Unknown No impact 
DOE Interdependent Infrastructure Systems Resilience Analysis for Enhanced Microreactor Power Grid Penetration Quantifies key resilience elements across integrated energy systems and their vulnerabilities to threats and hazards. This includes the ability to accurately analyze and visualize a region’s critical infrastructure systems ability to sustain impacts, maintain critical functionality, recover from disruptive events. Machine Learning (Type Unknown) Mapping Planning or development stage Indirect impact This advanced decision support capability can improve our understanding of these complex relationships and help predict the potential impacts that microreactors and distributed energy resources have on the reliability and resiliency of our energy systems. 
DOE Adaptive Fingerprinting of Control System Devices through Generative Adversarial Networks Reduces manual labor and operational cost required for training an electromagnetic (EM)-based anomaly detection system for legacy industrial control systems devices and Industrial Internet of Things. Unclear Organization & Efficiency Unknown No impact 
DOE Support Vector Analysis for Computational Risk Assessment, Decision-Making, and Vulnerability Discovery in Complex Systems Combines a support vector machine and PRA software to auto-detect system design vulnerabilities and find previously unseen issues, reduce human error, and reduce human costs. Support Vector Machines Monitoring or Detection Unknown No impact 
DOE Deep Reinforcement Learning and Decision Analytics for Integrated Energy Systems Manages distributed or tightly coupled multi-agent systems utilizing deep neural networks for automatic system representation, modeling, and end-to-end learning. Neural Networks Mapping Unknown No impact 
DOE Nuclear-Renewable-Storage Digital Twin: Enhancing Design, Dispatch, and Cyber Response of Integrated Energy Systems Develops a learning-based and digital twin enabled modeling and simulation framework for economic and resilient real-time decision-making of physicsinformed integrated energy systems (IES) operation. Unclear Forecasting & Prediction Unknown Indirect impact Learningbased algorithms will make real-time decisions upon detection of component contingencies caused by climate-induced or man-made extreme events, such as cyber-attacks or extreme weather, thereby mitigating their impacts through appropriate counter measures. 
DOE Automated Infrastructure & Dependency Detection via Satellite Imagery and Dependency Profiles Produces innovative and stateof-the-art image processing results that advance abilities to secure and defend national critical infrastructure. Automated Image Processing Security Unknown Indirect impact 
DOE Accelerated Nuclear Materials and Fuel Qualification by Adopting a First to Failure Approach Physics-based multi-scale modeling was coupled with deep, recursive, and transfer learning approaches to accelerate nuclear materials research and qualification of highentropy alloys. Neural Networks Research (Other) Unknown No impact 
DOE Evaluating thermal properties of advanced materials Helps elucidate thermophysical properties of a material from a single laser flash measurement. Machine Learning (Type Unknown) Research (Other) Unknown No impact The standard thermal diffusivity measurement technique laser flash is enhanced by modifying the traditional experimental set up and analyzing results with a machine learning based tool that includes a finite element model, a least-squares fitting algorithm and experimental data treatment algorithms. 
DOE Spectral Observation Convolutional Neural Network Analyzes collected radiation spectra using advanced, scalable deep learning by combining spectroscopic expertise with high performance computing. Neural Networks Research (Other) Unknown No impact This method was trained, tested, and operated on the International Space Station’s Spaceborne Computer-2 supercomputer, returning zero errors over the course of 100 training hours. 
DOE Passive Strain Measurements for Experiments in Radiation Environments Determines permanent strains induced by irradiation and extract critical parameters using modeling and simulation as well as machine learning algorithms. Machine Learning (Type Unknown) Research (Other) Short-term project or study No impact 
DOE Machine Learning Interatomic Potentials for Radiation Damage and Physical Properties in Model Fluorite Systems Studies the influence of radiation damage on physical properties of calcium fluoride and uranium dioxide. Machine Learning (Type Unknown) Research (Other) Short-term project or study No impact The high throughput capability of this method will become an important combinatorial materials science tool for developing and qualifying new nuclear fuels. 
DOE Data-driven failure diagnosis and prognosis of solid-state ceramic membrane reactor under harsh conditions using deep learning technology with internal voltage sensors Investigates in situ the effects of different components on the degradation behavior in a solid-state ceramic membrane reactor by embedding sensors that will collect current and impedance data during operation. Unclear Research (Other) Short-term project or study No impact 
DOE Tailoring the Properties of Multiphase Materials Through the Use of Correlative Microscopy and Machine Learning Identifies and correlates the critical microstructural features in a multiphase alloy that exhibits high strength and fracture toughness. Neural Networks Research (Other) Unknown No impact Experimental data will be used to train a convolutional neural network (CNN) in a semi-supervised environment to identify key microstructural features and correlate those features with the strength and toughness 
DOE Microstructurally-driven Framework for Optimization of In-core Materials Enables reactor developers to quickly understand the complex linkage between alloy composition, thermomechanical processing, the resulting microstructure, and swelling and creep behavior. Automated Image Processing Research (Other) Unknown No impact 
DOI Wildlife Underpass Camera Trap Image Classification, San Diego CA This software system takes wildlife camera trap images as inputs and outputs the probability of the image belonging to user-specified taxonomic classes based on wildlife species present in each image. Neural Networks Classification or Labeling Planning or development stage No impact The process of humans reviewing, labeling, and QA/QCing labels is labor intensive, time consuming, and costly. Developing AI systems that can perform these tasks within an acceptable level of accuracy can reduce the costs in extracting tabular data from camera-based datasets and increase the volume of data for analysis. 
DOI Walrus Haulout Camera Trap Image Classification Takes walrus haulout camera trap images as inputs and outputs the probability of the image containing walruses and various human disturbances (boats, aircraft, etc.). Neural Networks Classification or Labeling Planning or development stage No impact 
DOI ARMI Amphibian Species ID from Acoustic Data Takes audio clips that have been converted to sonograms (images) and classify the species generating the vocalizations in the recordings; initial prototype project will attempt to develop models that can identify audio clips containing bullfrog vocalizations. Neural Networks Classification or Labeling Planning or development stage No impact Reviewing audio recordings and identifying species vocalizations captured therein is time consuming and labor intensive. For these reasons, many recordings remain unprocessed, preventing valuable data from being available for analysis. 
DOI Individual Mountain Lion ID from Camera Data Takes pairs of mountain lion facial images and outputs the probability that the images come from the same individual mountain lion. This will allow researches to passively "mark" individuals and support population estimation analyses. Neural Networks Monitoring or Detection Planning or development stage No impact 
DOI Walrus Object Detection in Drone/Satelite Imagery Inputs drone imagery and outputs bounding boxes for individual walruses. Neural Networks Monitoring or Detection Planning or development stage No impact If successful, this will allow researchers with Alaska Science Center to count the numbers of walruses in drone imagery to support population research. 
DOI PRObability of Streamflow PERmanence Incorporates sparse streamflow observation data representing wet or dry stream conditions and gridded hydroclimatic explanatory data to predict the annual probability of streamflow permanence at 30-m (PROSPER Pacific Northwest) or 10-m (PROSPER Upper Missouri) resolution. High-Power Computing Forecasting & Prediction Ongoing project (in production more than a year) No impact 
DOI Water Mission Area Drought Prediction Project Predicts daily hydrologic drought using machine learning models calibrated on streamflow data (response) and meteorological forcing data. High-Power Computing Forecasting & Prediction Planning or development stage No impact 
DOI Water Mission Area Regional Drought Early Warning System Predicts and forecasts daily hydrologic drought in the Colorado River Basin (CRB). Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Planning or development stage No impact Uses gridded meteorologic forcing data and daily streamflow data in the CRB to build random forest and neural networks (long-short term memory) to determine the best approach to predicting and forecasting hydrologic drought. The project is being developed on AWS and in cooperation with CHS. We are also using the USGS HPC systems. 
DOI AI system to recognize individual fish and disease Recognizes individual fish and their disease status from images. Success of this effort could complement or replace traditional mark-recapture methods used for estimating abundance, survival, and movement, and this could greatly reduce costs to fisheries managers. Automated Image Processing Monitoring or Detection Ongoing project (in production less than a year) No impact 
DOI River Image SEnsing Development of a reliable camera system for integration into the operational streamgage monitoring network of the USGS Water Mission Area; capable of producing time-series of surface water levels derived from still camera images using AI/ML modeling techniques. Automated Image Processing Monitoring or Detection Ongoing project (in production less than a year) No impact 
DOI Estimating stream flow from images in headwaters Measures how much water flows in small, ungaged stream networks using timelapse images captured by inexpensive and off-the-shelf cameras and provides a web-based platform for making the images, associated climate and other related data as well as the model itself easy to access and explore. Automated Image Processing Monitoring or Detection Ongoing project (in production less than a year) No impact 
DOI Economic valuation of fisheries in the Delaware River Links existing hydrological flow data and models with trout population dynamic models, changes to fish catch, and the economic benefits of recreational fishing. Neural Networks Monitoring or Detection Ongoing project (in production less than six months) No impact 
DOI Stream physical habitat characterization in the Chesapeake Bay Watershed Takes a large dataset of rapid habitat assessment data collected by multiple jurisdictions in the Chesapeake Bay Watershed, train a predictive model using those data, and use that model to predict stream habitat conditions for all unmeasured stream reaches in the region. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Ongoing project (in production less than a year) No impact 
DOI Deep Learning for Automated Detection and Classification of Waterfowl, Seabirds, and other Wildlife from Digital Aerial Imagery Automates the detection of wildlife in aerial imagery and the taxonomic classification of wildlife from the binary detector. Automated Image Processing Monitoring or Detection Ongoing project (in production less than a year) No impact Uses Tallgrass to develop and train algorithms, BlackPearl/Caldera to store large image datasets, a hosted instance of a customized version of the Computer Vision Annotation Tool to gather manually annotated data, and a separate PostgreSQL database to store annotations and image metadata. 
DOI Prediction of Regolith Thickness in the Delaware River Basin Uses observations of the depth to bedrock reported by private well drillers in the Delaware River Basin to map the thickness of the regolith layer. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Ongoing project (in production less than a year) No impact 
DOI ML-Mondays course on applications of deep learning to image analysis A course in application of deep learning image segmentation, image classification, and object-in-image detection. Machine Learning (Type Unknown) Research (Other) Ongoing project (in production more than a year) No impact 
DOI Coast Train Dataset of orthomosaic and satellite images of coastal, estuarine, and wetland environments and corresponding thematic label masks. Automated Image Processing Classification or Labeling Ongoing project (in production less than six months) No impact The data consist of spatial and time-series, and contains 1.2 billion labelled pixels, representing over 3.6 million hectares. 
DOI Seabird and Marine Mammal Surveys Near Potential Renewable Energy Sites Offshore Central and Southern California Images output from the final model classified targets into seven categories: bird, dark bird, dark bird flying, light bird, fish, marine mammal, and other. Next, reclassify model labels to the lowest taxonomic group possible. Automated Image Processing Classification or Labeling Short-term project or study No impact The Seabird Studies Team at the Western Ecological Research Center (WERC), with support from the Bureau of Ocean Energy Management (BOEM), completed aerial photographic surveys of the ocean off central and southern California between 2018-2021. Over 800,000 high resolution images of the ocean were collected, with the goal of extracting and counting marine birds and mammals contained within. Once low taxonomic reclassification is complete, we will generate maps of species distribution and abundance to inform BOEM’s planning in advance of potential offshore wind energy development along the California coast. 
DOI Fouling Identification Neural Network (FINN) Predicts and detects sensor (sonde) fouling at USGS stream gages. Neural Networks Monitoring or Detection Ongoing project (in production more than a year) No impact 
DOI Mapping river bathymetry from remotely sensed data Uses high frequency satellite images from the Planetscope constellation to estimate water depth in river channels. Neural Networks Forecasting & Prediction Planning or development stage No impact The training data consist of field measurements of water depth collected as part of other USGS projects on five different rivers. The neural network regression method is implemented in MATLAB using the Deep Learning Toolbox. 
DOI Mapping benthic algae along the Buffalo National River from remotely sensed data Uses orthophotos acquired from a manned, fixed-wing aircraft and multispectral images from two different satellites to map bottom-attached (benthic) algae along the Buffalo National River in northern Arkansas. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Planning or development stage No impact 
DOI Characterization of Sub-surface drainage (tile drains) from satellite imagery Delineates tile drains in satellite imagery, providing a way to look at historical imagery and to use satellite data to maintain an up-to-date geospatial layer of tile drain extent in basins of interest. Automated Image Processing Mapping Ongoing project (time unknown) No impact Uses panchromatic imagery that is processed using a UNet model that was trained on a library of panchromatic images on which visible tile-drain networks had been traced; uses a combination of python scripting that is encapsulated in a Jupyter notebook. 
DOI Waterfowl Lifehistory and Behavior Classification Provides a highly accurate daily classification of waterfowl behavior into 8 life history states/movement patterns using hourly GPS relocations and, optionally, remotely sensed habitat data. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Classification or Labeling Planning or development stage No impact 
DOI Spot Elevation OCR from historical topo maps Creates a database of summit spot elevations from the HTMC labeled for summits in CONUS. Optical Character Recognition (or Text Extraction) Classification or Labeling Ongoing project (in production more than a year) No impact 
DOI TerrainFeatures detection and recognition Uses DL tools to extract terrain features. Neural Networks Monitoring or Detection Ongoing project (in production more than a year) No impact 
DOI The National Landcover database Develops Landcover across all 50 states; includes HPC processes, cloud services, and local resources to create thematic and continuous field classifications. Machine Learning (Type Unknown) Classification or Labeling Ongoing project (time unknown) No impact These classifications serve as the base for users and federal agencies across the nation to provide wildlife habitat estimates, urban runoff estimates, population growth, etc. 
DOI Artificial Intelligence for Environment & Sustainability (ARIES) ARIES is a full-stack solution for integrated modelling, supporting the production, curation, linking and deployment of scientific artifacts such as datasets, data services, modular model components and distributed computational services. This design enables automation of a wide range of modeling tasks that would normally require human experts to perform. Network Analysis (ie Bayesian, or Social Network) Mapping Ongoing project (in production more than a year) No impact ARIES is an international research project based at the Basque Centre for Climate Change (Bilbao, Spain), to which USGS has been a long-term collaborator. ARIES uses semantics and machine reasoning to enable AI-assisted multidisciplinary, integrated modeling of coupled human-natural systems. 
DOI Global Inland Fisheries Risk Index Informs the relative influence of threats in the development of a global inland fisheries assessment using boosted regression trees to derive a spatially-explicit risk index of stressors. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Ongoing project (in production more than a year) No impact 
DOI Fish and Climate Change Database (FiCli) Automates certain portions of the review process to increase efficiency in maintaining and updating the database. Natural Language Processing Organization & Efficiency Ongoing project (in production less than a year) No impact The Fish and Climate Change Database (FiCli) is a comprehensive database of peer-reviewed literature compiled through an extensive, systematic primary literature review to identify English-language, peer-reviewed journal publications with projected and documented examples of climate change impacts on inland fishes globally. 
DOI Evaluating fish movement in restored coastal wetlands using imaging sonar and machine learning models Wetland managers are restoring coastal wetland habitats in the Great Lakes, and often seek more information on when and how fish access restored habitats. Terabytes of hydroacoustic data on fish movement need to be analyzed more efficiently, so a collaboration between USGS, USFWS, and the University of Michigan is developing a machine learning model (MLM) that identifies, tracks, and quantifies fish movement. Neural Networks Monitoring or Detection Ongoing project (in production less than a year) No impact The completed model will read proprietary sonar image files, convert them to a universal file format (i.e., .mp4), place bounding boxes around individual fish detected by the model, and track them across consecutive image frames to determine bi-directional movement. The model uses training data and TensorFlow-based convolutional neural networks for object detection. 
DOI Fluvial Fish Native Distributions for the Conterminous United States using the NHDPlusV2.1 and Boosted Regression Tree (BRT) Models Develops species distribution models for 271 fluvial fish species in their native ranges of the conterminous United States. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Ongoing project (in production less than six months) No impact 
DOI Prediction of Inland Salinity in the Delaware River Basin Inputs watershed characteristics (soils, land cover), land use (road salt application) and meteorological timeseries, and output predictions of specific conductance (SC) for inland stream reaches in the Delaware River Basin (DRB). Neural Networks Forecasting & Prediction Planning or development stage No impact The model will be trained using SC sample data from within the DRB. The resulting model will allow for predictions in ungaged locations and time periods, and allow for an evaluation of salinity exposure in these stream reaches. The model will be built using pyTorch on the USGS Tallgrass supercomputer. 
DOI Prediction of Salt Front Location in the Delaware River Estuary Makes predictions of the 250 mg/L isochlor (salt front location) within the Delaware River Estuary. The model will be driven by river discharge into the estuary, tidal forcings, and meterological data from several points throughout the estuary. Neural Networks Forecasting & Prediction Planning or development stage No impact Model predictions will be compared with a process-based, hydrodynamic model, COAWST. 
DOI Prediction of Water Temperature in the Delaware River Basin Makes water temperature predictions at 456 reaches in the Delaware River Basin. Neural Networks Forecasting & Prediction Ongoing project (in production more than a year) No impact The recurrent graph convolutional network (RGCN) was pre-trained with predictions from a coupled process-based model that predicts stream flow and temperature. 
DOI Forecasting Water Temperature in the Delaware River Basin Produces 7-day forecasts of daily maximum stream water temperature downstream of drinking water reservoirs in support of water management decisions. Neural Networks Forecasting & Prediction Ongoing project (in production less than a year) Indirect impact Our process-guided deep learning model was pretrained on output from an integrated stream-reservoir process-based model and used an autoregressive technique and data assimilation to ingest real-time observations of stream temperature to improve near-term forecasts. 
DOI Prediction of Flood Flow Metrics for Minimally Altered Catchments Inputs watershed characteristics (soils, land cover) and long-term meteorological data, and outputs predictions of flood flow metrics (magnitude, duration, frequency, volume) for stream reaches. Machine Learning (Type Unknown) Forecasting & Prediction Planning or development stage No impact The resulting models will allow for estimating flood flow metrics in ungaged reaches, which can be used to inform infrastructure designs along those reaches (e.g., bridges). 
DOI Process-Guided Deep Learning Predictions of Lake Water Temperature Predicts depth-specific lake temperatures while obeying physical laws using inputs of meteorological drivers. Neural Networks Forecasting & Prediction Ongoing project (in production more than a year) No impact 
DOI Prediction of Lake Water Temperature using Lake Attributes Inputs lake characteristics (surface area, elevation, and others to be determined) and outputs predictions of depth-specific lake temperatures. Neural Networks Forecasting & Prediction Planning or development stage No impact The models will be developed using various Python packages including PyTorch on the USGS Tallgrass supercomputer. 
DOI Process-Guided Deep Learning for Dissolved Oxygen Predictions on Stream Networks Predicts daily minimum, mean, and maximum dissolved oxygen (DO) concentrations at several stream locations in the Delaware River Basin. Neural Networks Forecasting & Prediction Planning or development stage No impact The deep learning models were written via TensorFlow, the data prepartion is in R, and the modeling workflow was scripted via Snakemake. 
DOI Multi-task deep learning for daily streamflow and water temperature Predicts two interdependent variables, daily average streamflow and daily average stream water temperature, together using multi-task deep learning. Neural Networks Forecasting & Prediction Planning or development stage No impact The stream temperature data were collected by the USGS and made available via NWIS. The streamflow observations were also collected by the USGS but collated along with input drivers in the CAMELS dataset. 3) This work was done using Python. The deep learning models were written via TensorFlow and the modeling workflow was scripted via Snakemake. 
DOI Predicting Water Temperature Dynamics of Unmonitored Lakes With Meta‐Transfer Learning Compares the transfer of different model types from well-observed to unobserved lake systems. Neural Networks Forecasting & Prediction Ongoing project (in production more than a year) No impact Process-based models, neural networks, and process-guided neural networks are trained on well observed lakes (source lakes) and then is used to make predictions in unobserved lakes (target lakes). 
DOI Process-guided deep learning for predicting stream temperature in out-of-bound conditions Predicts network wide daily average stream temperature in the Delaware River Basin; compares the performance of two deep learning achictectures, both of which incorporate process guidance through pretraining on process-based modelling outputs. Neural Networks Forecasting & Prediction Planning or development stage No impact 
DOI Process guidance for learning groundwater influence on stream temperature predictions Predicts network wide daily average stream temperature in the Delaware River Basin; focuses on developing a custom loss function that helps deep learning models learn to account for groundwater influence on stream temperature. Neural Networks Forecasting & Prediction Planning or development stage No impact 
DOI Explainable AI and interpretable machine learning Develops expertise and resources for Explainable AI (XAI) within WMA PUMP Projects. The inputs are various models developed for predicting stream temperature, discharge, dissolved oxygen, and other characteristics. The outputs are interpretable metrics to help understand why models are making the predictions they are and what physical processes are getting captured with the model architectures. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Research (Other) Planning or development stage No impact 
DOI AI applications to mapping surface water Investigates the use of hand annotated hydrography from one region to train an artificial neural net (ANN) to identify where surface water is likely to be in other areas. Neural Networks Monitoring or Detection Planning or development stage No impact 
DOI Where’s the Rock: Using Neural Networks to Improve Land Cover Classification Differentiates exposed bare rock (rock) from soil cover (other) in order to classify bare rock in NAIP orthoimagery, starting with the Sierras, in order to provide a more accurate map of soil vs. rock-covered areas for use in landslide hazard mapping, quantifying soil carbon storage, calculating water fluxes, etc. Neural Networks Classification or Labeling Ongoing project (in production more than a year) No impact 
DOI Data–driven prospectivity modelling of sediment–hosted Zn–Pb mineral systems and their critical raw materials Produces a prospecticvity map for Clastic Dominated and Mississippi Valley Type deposits in the three countries. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Ongoing project (in production less than a year) No impact 
DOI Updating Real-time Earthquake Shaking, Ground Failure, and Impact products with remote sensing and ground truth observations Enables accurate and high-resolution multi-hazard and damage estimates by jointly inferring shaking and secondary hazards and resulting building damage and quantifying their causal dependencies from imagery and prior loss and GF models. Network Analysis (ie Bayesian, or Social Network) Forecasting & Prediction Planning or development stage No impact The underlying physical causal dependencies are modeled using a multi-layer causal Bayesian network. Initial results are impressive, showing that our framework significantly improves the GF prediction abilities. 
DOI Using Artificial Neural Networks to Improve Earthquake Ground-Motion Models Provides estimates of peak ground-motion from earthquakes given the location, magnitude, and local geological structure at a site of interest. Neural Networks Forecasting & Prediction Planning or development stage No impact 
DOI Leveraging Deep Learning to Improve Earthquake Monitoring Characterizes earthquake source information using small portions of waveform data to improve autotmatic phase picking, classify phase types, and estimate source-station distances. Neural Networks Monitoring or Detection Ongoing project (in production less than a year) No impact 
DOI Using Gradient Boosting Method and Feature Selection to Reduce Aleatory Uncertainty of Earthquake Ground-Motion Models Develops ground-motion models for peak ground acceleration and peak ground velocity using a gradient boosting method (GBM). Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Mapping Planning or development stage Indirect impact In total 128 GBM-based ground-motion models are developed for estimating PGA and PGV, respectively, using varying subsets of explanatory variables. 
DOI Application of machine learning to ground motion-based earthquake early warning Predicts what the earthquake peak ground shaking will be across a region. Machine Learning (Type Unknown) Forecasting & Prediction Planning or development stage Indirect impact 
DOI A machine learning approach to developing ground motion models from simulated ground motions Build a ground motion model (GMM) from a synthetic database of ground motions extracted from the Southern California CyberShake study. An artificial neural network is used to find the optimal weights that best fit the target data (without overfitting), with input parameters chosen to match that of state-of-the-art GMMs. Neural Networks Forecasting & Prediction Ongoing project (in production more than a year) No impact 
DOI Integrating machine learning phase pickers into the Southern California Seismic Network earthquake catalog Evaluates the readiness of machine-learning models for automatic earthquake detection and phase picking to enhance the Southern California Seismic Network earthquake catalog, with the end-goal of using these models in routine seismic network operations. Neural Networks Monitoring or Detection Planning or development stage No impact 
DOI Understanding the 2020-2021 Puerto Rico Earthquake sequence with deep learning approaches Enhances the earthquake catalog for the 2020-2021 southwestern Puerto Rico earthquake sequence with a variety of deep learning approaches to understand its complex fault system, triggering mechanisms, and long-lived vigorous nature of the aftershock sequence. Neural Networks Research (Other) Short-term project or study No impact 
DOI Land Use Plan Document and Data Mining and Analysis R&D Explores the potential to identify patterns, rule alignment or conflicts, discovery, and mapping of geo history and/or rules. Inputs included unstructured planning documents. Outputs identify conflicts in resource management planning rules with proposed action locations requiring exclusion, restrictions, or stipluations as defined in the planning documents. Natural Language Processing Research (Other) Planning or development stage No impact 
DOI Data Driven Sub-Seasonal Forecasting of Temperature and Precipitation Deployed data driven methods for sub-seasonal (2-6 weeks into future) prediction of temperature and precipitation across the western US. Decision Tree Analysis (ie Random Forest or Gradient-Boosting) Forecasting & Prediction Planning or development stage No impact Improving sub-seasonal forecasts has significant potential to enhance water management outcomes. 
DOI Data Driven Streamflow Forecasting A year-long evaluation of existing 10-day streamflow foreasting technologies and a companion prize competition open to the public, also focused on 10-day streamflow forecasts. Forecasts were issued every day for a year and verified agains observed flows. Neural Networks Forecasting & Prediction Planning or development stage No impact Across locations and metrics, the top perfoming foreacst product was a private, AI/ML forecasting company - UpstreamTech. Several competitors from the prize competition also performed strongly; outperforming benchmark forecasts from NOAA. Reclamation is working to further evaluate the UpstreamTech forecast products and also the top performers from the prize competition. 
DOI Seasonal/Temporary Wetland/Floodplain Delineation using Remote Sensing and Deep Learning Provides improved seasonal/temporary wetland/floodplain delineation when high temporal and spatial resolution remote sensing data is available to inform the management of protected species and provide critical information to decision-makers during scenario analysis for operations and planning. Neural Networks Mapping Short-term project or study No impact 
DOI Improving UAS-derived photogrammetric data and analysis accuracy and confidence for high-resolution data sets using artificial intelligence and machine learning UAS derived photogrammetric products contain a large amount of potential information that can be less accurate than required for analysis and time consuming to analyze manually; apply machine learning to better analyze photogrammetric products. Machine Learning (Type Unknown) Research (Other) Planning or development stage No impact 
DOI Photogrammetric Data Set Crack Mapping Technology Search Explores a specific application of photogrammetric products to process analysis of crack mapping on Reclamation facilites. Unclear Mapping Planning or development stage No impact 
DOI Improved Processing and Analysis of Test and Operating Data from Rotating Machines This project is exploring a better method to analyze DC ramp test data from rotating machines. Previous DC ramp test analysis requires engineering expertise to recognize characteristic curves from DC ramp test plots. The ramp test plots can be analyzed by computer software, rather than manual engineering analysis, to recognize characteristic curves. Regression Analysis Research (Other) Planning or development stage No impact 
DOI Sustained Casing Pressure Identification Quickly identify wells with sustained casing pressures to mitigate accidents on well platforms. Neural Networks Monitoring or Detection Planning or development stage No impact 
DOI Level 1 Report Corrosion Level Classification Automated screening system that can identify parts of wells that exhibit excess corrosion to greatly reduce report processing time. Neural Networks Organization & Efficiency Planning or development stage No impact 
DOI Well Activity Report Classification Researches use of self-supervised deep neural networks to identify classification systems for significant well event using data from well Activity Reports. Neural Networks Research (Other) Planning or development stage No impact