2026-09-03 –, Phoenix Hall
As AI agents and LLM-based systems become primary consumers of geospatial data, the weak point in most data infrastructure is no longer access — it's meaning. Open data alone doesn't guarantee interoperability; without a shared semantic layer, every new API becomes another silo an AI agent has to be taught to navigate. This talk shares Geolonia's experience building GeonicDB, a context data platform built on the NGSI-LD standard — including why we chose not to simply adopt an existing broker like FIWARE Orion-LD, and built our own instead. We'll cover the reasoning, the trade-offs, and a live demo of AI agents querying standardized context data directly.
Local governments and organizations have spent the last decade opening up their data, yet most "open data" portals remain difficult for both humans and machines to use meaningfully — inconsistent schemas, undocumented fields, and API sprawl are the norm. This problem is about to get worse, not better, as AI agents start querying data infrastructure directly and autonomously, without a human in the loop to interpret ambiguity.
This talk makes the case that the next phase of geospatial data infrastructure needs to be built around open, machine-interpretable standards for context and meaning — not just open licenses. We'll draw on Geolonia's experience designing GeonicDB, a context data management platform built on NGSI-LD (the ETSI/FIWARE standard for context information), to explore:
- Why a standards-based, entity-relationship semantic model matters more than a conventional REST/GeoJSON API approach for AI-era data consumption
- Why we didn't simply adopt FIWARE Orion-LD, an existing NGSI-LD context broker — the specific gaps that led us to build GeonicDB instead, and what that decision cost and bought us
- How a shared semantic layer changes what's possible for AI agents and RAG systems querying geospatial and civic data — improved retrieval accuracy, fewer hallucinations from ambiguous schema, and more reliable tool-calling
We'll close with a short live demo showing an AI agent querying standardized context data, and open questions for the FOSS4G community on where standards-based interoperability should go next in the agentic AI era.
This talk is aimed at practitioners working on geospatial data platforms, open data infrastructure, and anyone thinking about how AI will consume (not just display) geospatial data.
Geolonia Inc.
Level of technical complexity: 2 - intermediate Indicate what is (are) the open source project(s) essential in your talk:NGSI-LD, FIWARE Orion-LD
GeonicDB, our context data platform discussed in this talk, is built on these standards/projects. GeonicDB itself is not yet open source but is in preparation for open-sourcing.
Hal Seki is CAIO and Director at Geolonia, a Japanese geospatial technology company, where he leads AI strategy and the development of GeonicDB, a context data platform for civic and geospatial applications. He is also the founder of Code for Japan, a civic tech nonprofit, and serves as a Senior Expert at Japan's Digital Agency. His work spans open data, geospatial infrastructure, and AI-driven public sector systems, with a focus on making data usable — for both people and machines — across local government and civic contexts.