2025 FOSS4G NA | Scalable GeoAI - Nathan McEachen
In this session from FOSS4G NA 2025, Nathan McEachen, founder and CEO of TerraFrame, presents Scalable GeoAI: Harnessing Agents, Knowledge Graphs, and DGGS. Nathan discusses the critical need for scalable, real-time geospatial intelligence in disaster response and resiliency contexts, moving beyond simple chatbots to systems that understand the semantic relationships between geographical features and their downstream impacts.
The presentation details a multi-agent architecture that leverages Spatial Knowledge Graphs (SKGs) to model complex relationships and Discrete Global Grid Systems (DGGS) to normalize and aggregate statistical data. Nathan demonstrates how these technologies work together to provide traceable, deterministic answers to natural language questions—such as identifying which census areas are compromised by power failures during a flood—while emphasizing the importance of semantic identity and interoperable data publishing standards to avoid creating new information silos.
Highlights:
🛰️ The Scalability Challenge: Why AI alone can't magically integrate spatial data and the need for upfront interoperability in National Spatial Data Infrastructures
🔄 Discrete Global Grid Systems (DGGS): An introduction to the emerging OGC standard that quantizes the world into uniform zones for efficient data normalization
🧩 Spatial Knowledge Graphs: Moving from feature-type centric tables to relationship-centric graphs that Large Language Models can traverse and understand
⚙️ Agentic AI Tools: Using the Model Context Protocol (MCP) to expose data endpoints such as DGGS servers and triple stores to autonomous AI agents
🌐 Disaster Resiliency Demo: Watch a real-world prototype query flood levels near Winnipeg and automatically correlate power infrastructure with census boundaries
🏢 Beyond Silos: The vision for "No STAC"—transforming SpatioTemporal Asset Catalogs into interoperable knowledge graphs for cross-agency data discovery
🏗️ Avoiding Monoliths: Why the goal should be a standardized approach for publishing federated graphs rather than "one big graph to rule them all"
For more content like this check out www.projectgeospatial.com
#Geospatial #FOSS4G #GeoAI #KnowledgeGraphs #DGGS #DisasterResiliency #AI #OpenSource #DataEngineering #STAC #OGC #TerraFrame #ProjectGeospatial #MachineLearning #SemanticWeb
