Twelve teams led by innovators from industry and academia will receive a total of $1 million in AWS credits to advance cutting-edge geospatial generative AI projects
St. Louis, MO — The Taylor Geospatial Institute (TGI) and Amazon Web Services (AWS) today unveiled a list of twelve innovative project teams that will share in $1 million in AWS credits to support development of new geospatial applications for generative artificial intelligence (AI). The winners were selected from 55 proposals submitted by teams including more than 200 accomplished geospatial scientists and practitioners from five continents.
Formally launched on October 29, 2024, in St. Louis, the Generative AI for Geospatial Challenge brought together leading minds in geospatial technology, AI, and cloud computing. Participants were invited to transform visionary ideas into actionable projects that could shape the future of geospatial innovation. Offering up to $1 million in AWS credits, the initiative empowered participants to focus on creativity and problem-solving while minimizing resource constraints.
The Challenge sparked a flood of interest, generating proposals on topics as varied as dialysis availability for end-stage renal disease patients to a system for near-real-time wildfire recovery planning. The winning proposals emerged after detailed review, scoring, and deliberation by a selection committee comprising six geospatial and AI experts, three each from AWS and TGI.
“We see and hear every day about the power and possibility of generative AI technology, but its promise has not yet been fully realized in the geospatial world,” said TGI Executive Director Nadine Alameh. “Through this challenge, TGI and AWS have drawn together experts from industry and academia with novel concepts for geospatial uses of generative AI and we have chosen twelve projects that will advance science and practice and solve hard problems in the real world.”
Through this challenge, TGI and AWS have drawn together experts from industry and academia with novel concepts for geospatial uses of generative AI and we have chosen twelve projects that will advance science and practice and solve hard problems in the real world.
Winning proposal teams came from a variety of backgrounds. Some were led by commercial companies with substantial involvement and support from academic scientists. Others were scholar-led with a multi-institution team and some support from an industry partner. Still other awarded projects were the proposals of a single company, or were entirely academic.
The varied teams yielded a broad range of novel project types and domains. Winning proposals include an effort to improve biodiversity net gain assessments, a system for real-time monitoring of deforestation in the Amazon, a tool for understanding the feasibility and environmental benefits of geothermal heat pump installations on residential or commercial sites, and AI-driven projections of future land cover changes.
Awardees are expected to use their awarded credits for projects to be completed in April 2025, with progress tracked by TGI’s GeoAI Working Group followed by opportunities for showcasing public demonstrations later in 2025.
Information about each awardee is below. More information on the Challenge is available here: https://taylorgeospatial.org/awschallenge/
Generative AI for Geospatial Challenge Awardees
(Alphabetical by Lead Proposer name)
Lead: Hamed Alemohammad
Proposal Team: Clark University CGA Team (Hamed Alemohammad — Faculty Member, Lead Proposer, Sam Khallaghi — Postdoctoral Researcher), Washington University in St. Louis Team (TGI Consortium Institution, Nathan Jacobs — Faculty Member, Srikumar Sastry — PhD Student, Dan Cher — PhD Student)
Project Synopsis/Abstract
This project leverages generative AI models to project future land cover changes using historical satellite imagery and land cover maps. By exploring three novel AI modeling approaches, we aim to enhance the accuracy in predicting environmental transformations. The project addresses critical geospatial challenges by supporting sustainable land-use planning and climate resilience strategies. Anticipated outcomes include scalable methodologies for integrating geospatial data and generative AI, fostering cross-sectoral applications in conservation, urban development, and policymaking, ultimately advancing the understanding and management of dynamic landscapes.
Lead: Avineon
Proposal Team: Avineon, Inc. (Darryl Murdock and Oliver Morris), Midgard Raven (Adam Simmons), Analytic Folk (Chul Gwon), Bunting Labs (Brendan Ashworth and Michael Egan), Earth Fire Alliance (EFA), Heavy.ai
Project Synopsis/Abstract
Problem Statement: Wildfires necessitate rapid, coordinated responses. However, local, state, federal, and international organizations struggle with real-time information sharing, hindering data integration, damage analysis, and resource allocation. The absence of a unified near-real-time geospatially enabled platform for assessing human impact impedes effective wildfire recovery efforts, delaying crucial actions and exacerbating the long-term impacts of wildfires. Solution: We will build and train a geospatially enabled system that uses fire models, incident management reports, and remotely sensed thermal data to provide near-real-time wildfire recovery plans.
Lead: Bedrock Research
Proposal Team: Bedrock Research LLC (Dr. Matt Reisman, Kevin LaTourette), AgileView, Inc. (Avi Lindenbaum, Peter Shagnea), Multimodal Vision Research Lab @ Washington University (Prof. Nathan Jacobs — TGI researcher)
Project Synopsis/Abstract:
Under TGI, the Bedrock Research team aims to build cross-modal remote sensing foundation models between synthetic data and diverse real GEOINT imaging modalities for rapid adaptation to rare/novel targets and applications. Bedrock’s deep mission knowledge and operational AI expertise combines with AgileView’s synthetic data generation techniques and cutting-edge academic research in computer vision from Prof. Nathan Jacobs’ Multimodal Vision Research Lab at Wash U to help usher in this new paradigm of generative AI-driven image processing capabilities. Our goal is to significantly lower the cost and timeline of building targeted computer vision models for defense and broad commercial applications.
Lead: Crosswalk Labs
Proposal Team: Crosswalk Labs (Daniel M Sheehan, Jason Burnett , Dr. Phil DeCola, Dr. Victoria Hunt, Dr. Geoff Roest, Dr. Anastasia Montgomery)
Project Synopsis/Abstract:
Accurate modeling of building asset-level emissions requires detailed geospatial and attribute data, yet datasets like the National Structures Inventory (NSI) often suffer from missing, incomplete, or erroneous information. This project leverages Generative AI services in Amazon Bedrock and serverless architecture (e.g., API Gateway, Lambda) to create intelligent agents that verify and enrich building property data. These agents will fill data gaps, validate attributes, and correct outliers, enabling the development of high-quality emissions models. The result will be a comprehensive and scalable approach to assessing building-level emissions across the United States.
Lead: Danti
Proposal Team: Danti (Martice Nicks, CTO, Dr. Anthony Hylick, Head of Engineering and AI, Dr. Charlie Veal, Sr AI Engineer)
Project Synopsis/Abstract
In this project, we will integrate feedback from FEMA, USGS, and other agencies involved in rapid response to natural disasters in the United States. These organizations invest significant time and effort in sourcing multimodal information from social media platforms, correlating imagery from airborne and space-based sensors, and analyzing government reports to conduct swift damage triage and assessments. Danti will enhance its multimodal AI system by incorporating large language models (LLMs) and other advanced tools to streamline these analytical workflows. This innovation will expedite disaster declarations, enabling states to access federal funding more quickly and ensuring citizens receive timely support.
Lead: Deep Earth
Proposal Team: Deep Earth (Christie Capper, Technical Lead, Albi Wiedersberg, Product & Partnerships Lead, Hamid Omidvar, Geospatial ML Engineer), Taylor Geospatial Institute (Yu-Feng Forrest Lin, Director, Illinois Water Resources Center, University of Illinois Urbana-Champaign, Andrew Stumpf, Principal Research Scientist, Illinois State Geological Survey, University of Illinois Urbana-Champaign, Ashlynn Stillwell, Associate Professor, Civil and Environmental Engineering, The Grainger College of Engineering, University of Illinois Urbana-Champaign, Jonathon G.W. Callicoat, Adjunct Instructor Geothermal Heat Transfer Systems, Saint Louis University) Technical Advisor (Alexander Van Noort, Geothermal and heat energy solutions, Ennatuurlijk Aardwarmte, Geospatial, geothermal and heat pump service company)
Project Synopsis/Abstract:
This project develops Thermal Atlas, an AI-powered tool to assess the feasibility, cost-effectiveness, and environmental benefits of geothermal heat pump (GHP) installations for residential and small commercial properties. We aim to accelerate GHP adoption and energy transition goals, reducing carbon emissions in key markets such as Europe and the United States. Upon entering an address, Thermal Atlas leverages generative AI to analyze geospatial, subsurface, building, and other data to generate personalized reports evaluating GHP suitability, installation costs, and energy savings. It features a user-friendly interface with 3D visualizations and conversational AI to simplify decision-making and connect users to vetted providers.
Lead: Earth Genome
Proposal Team: Earth Genome (Ben Strong, Edward Boyda, Hutch Ingold, Mikel Maron)
Project Synopsis/Abstract:
Earth Index uses large geospatial AI foundation models to simply and rapidly identify critical places of interest and accelerate environmental monitoring. Earth Index powers food systems applications like identifying cattle feeding operations globally and monitoring impact of tropical commodities production on deforestation. This project will fuse multiple novel data sources as input to geospatially trained AI foundation models, enabling wider set of food system security applications by allowing for localized fine tuning of these generative AI models in a streamlined and scalable workflow, via an extended API integrated with remote sensing scientists’ workflows in notebooks, QGIS and other tools.
Lead: Sean Gorman
Proposal Team: Zephr.xyz Inc., Pramukta Rao PhD, Sean Gorman PhD, Kostas Stamitou PhD, Joseph Strus PhD.
Project Synopsis/Abstract:
Zephr is proposing a system to location enable generative AI. Large language models (LLM’s) can struggle with understanding geographic relationships with “place”. To overcome this we can explicitly detect geographic entities in a user’s view and connect that place to a LLM for exploration. This will explicitly interconnect “place” in the real world with the aggregated knowledge about places embedded in a LLM. The combination will open up the near term possibility of audio AI assistants being able to explicitly connect to places through location.
Lead: LuxCarta International
Proposal Team: Sacha Lepretre, LuxCarta AI Team
Project Synopsis/Abstract:
LxGenAIEarthMapper — We propose a generative AI system capable of making sense of the world viewed from above with the ability to interactively adapt itself to end user needs. Leveraging the latest AI technologies such as LLMs (Large Language Models) and VLMs (Vision Language Models), we are developing an AI assistant capable of generating any kind of map from satellite imagery augmented with the semantics requested by the user. Our AI system allows the end user to directly interact with the raw imagery and extract exactly what they want from it in an interactive and fast manner, with no other middleman. LxGenAIEarthMapper will be part of the LuxCarta’s BrightEarth.ai AI platform to generate on-demand 3D geospatial data from anywhere on the planet’s surface.
Lead: Sparkgeo Consulting Inc.
Proposal Team: Sparkgeo (James Banting is the Vice President of Research and Strategic Projects), UK Multimodal AI Network (The Alan Turing Institute), University of Cambridge (Dr. Nataliya Tkachenko)
Project Synopsis/Abstract
This project applies Generative AI to improve Biodiversity Net Gain (BNG) assessments, a UK regulatory requirement ensuring land development provides measurable environmental benefits. By fine-tuning geospatial models, it automates habitat mapping, biodiversity metrics, and scenario planning using satellite imagery, aerial surveys, and local records. A digital assistant powered by large language models delivers insights through natural language queries and dashboards. This solution offers scalable tools for sustainable development and regulatory compliance while emphasizing BNG’s role as an economic indicator influencing investments in physical assets.
Lead: Tera Analytics
Proposal Team: Robert Carroll, Charles Mondello
Project Synopsis/Abstract
Tera GeoGAN is a cutting-edge AI platform leveraging the Pix2Pix model of Generative Adversarial Networks (GANs) and vector databases to analyze pre- and post-disaster imagery. By utilizing the Discriminator Model of Pix2Pix, it enhances inference accuracy, models complex damage patterns, and predicts future risks with precision. The platform integrates geospatial data, including satellite sensors, airborne surveys, and GIS data, within a vector database of damage assessments for rapid and reliable analytics. Additionally, it will use these diverse datasets to generate a Risk Index factor for properties based on post-disaster models, combined with datasets such as historical weather patterns. This approach provides high-fidelity insights for public safety, disaster planning, and insurance, enabling stakeholders to make informed decisions and expedite disaster response efforts.
Lead: Dr. Dong Xu
Proposal Team: Dong Xu (faculty member at TGI consortium university, University of Missouri-Columbia) Joe Soundararajan (TGI Consortium university student), Through Sensing LLC (Andrew Kalukin), Soundararajan Ezekiel (Indiana University of Pennsylvania)
Project Synopsis/Abstract
This project will develop an AI-driven system for near real-time monitoring of deforestation and biodiversity changes in the Amazon rainforest using multi-sensor satellite imagery. By integrating data from Sentinel‑1, Sentinel‑2, and Landsat satellites with advanced generative AI techniques, we generate high-resolution, spatially explicit estimates of deforestation extent, timing, patterns, and associated biodiversity impacts across the entire Amazon biome. Our approach leverages a novel hybrid architecture combining ResNet for spatial feature extraction and Temporal Convolutional Neural Networks or Long Short-Term Memory networks for time series analysis. Our solution aims to revolutionize conservation planning across the nine Amazon countries.
About Taylor Geospatial Institute
TGI is passionate about fueling geospatial science and technology to create the next generation of solutions and policies that the whole world will depend on for sustainability and growth.
The TGI consortium includes Saint Louis University, the Donald Danforth Plant Science Center, Harris-Stowe State University, University of Illinois Urbana-Champaign, Missouri University of Science & Technology, University of Missouri-Columbia, University of Missouri-St. Louis, and Washington University in St. Louis. Collectively, these institutions cover geospatial research from ocean depths to outer space.
For more information, visit taylorgeospatial.org.