There is an enormous amount of public geospatial data out there, and much of it describes the same things over and over — just never the same way twice. Different agencies publish the same kind of feature under different schemas, different field names, different coded values, and varying data quality. The data is free and open, but before you can actually use these sources together, you have to make them agree with one another and conform to a single, shared data model. That's the hard part, and it's rarely as simple as a download.

This talk is a hands-on demo of a workflow for doing exactly that, with AI doing the heavy lifting where it genuinely helps. We'll walk the workflow end to end:

• Finding sources. How we size up candidate public datasets before a single byte is loaded.
• Browse and select. A simple app to explore what's available and choose which sources to process — a service viewer and a catalog to pick from.
• AI-assisted field matching. Mapping each source's fields onto a shared data model, where semantically identical fields almost never share a name.
• Value editing. Reconciling inconsistent coded values and conventions so records from different publishers become directly comparable.
• Conversion. Sources are converted through a fast, storage-backed pipeline so the process stays quick and repeatable at scale.

The bulk of the session is the live demo. We'll take a familiar, everyday feature — think a common "roads" or street layer, published by several different sources — and carry it from raw, mismatched services to a single unified layer. Along the way we'll be candid about where the AI accelerates the work, where it gets things wrong, and where a human still has to stay in the loop.

Who should attend: GIS practitioners, data engineers, and analysts who work with public or open geospatial data and need to make heterogeneous sources agree with one another.