What does it actually take to make AI useful on location data?
In this episode, Matt Forrest sits down with Ryan Urabe, co-founder and CTO of Dataplor, to unpack how AI, embeddings, and agents are changing the way we work with points of interest and places data.
Ryan explains why general-purpose models already understand spatial concepts but still struggle to execute them, and why the real unlock is the harness around the model, not a geospatial-specific model. He walks through Dataplor's data-quality philosophy, the category problem (why "supermarket" and "grocery store" have zero string similarity but near-zero conceptual distance), and how embeddings let them measure conceptual distance across 10^9 places and even across languages.
Whether you build with spatial data, lead a data team adopting AI, or you are trying to figure out what embeddings actually do, this conversation maps out what is working today and what is still forming.
In this episode, we cover:
- Why AI is an accelerant for data quality, not a replacement for it
- Treating AI like a capable employee on their first day
- Where general models fall short on spatial problems
- DuckDB as the Swiss Army knife for orchestrating spatial data
- The category problem and conceptual vs. semantic distance
- How embeddings map 7-Eleven in Tokyo and Tennessee to the same concept
- A vision for agentic AI built natively for geospatial
- The "end of the scarcity of intelligence" framing for where this is all heading
Connect with Ryan:
LinkedIn:
https://www.linkedin.com/in/rurabe/Website:
https://www.dataplor.comEmail:
ryan@dataplor.comLEARN MORE
Dataplor's agentic SaaS product is launching this summer. To start your complimentary trial, contact:
freetrial@dataplor.com📊 FREE: The Modern GIS Skill Map
The 5 skills that actually matter in modern GIS (and what you can stop learning). Based on a survey of 1,400+ geospatial professionals.
➡ Get the free training + PDF guide:
https://forrest.nyc/go/training/00:00:00 – Cold open
00:01:01 – Welcome and Ryan's background
00:03:44 – Why AI still struggles with location
00:05:13 – Bringing Dataplor's data into AI, product and team
00:08:36 – How technical teams are adopting AI
00:10:10 – Treating AI like a capable new employee
00:12:20 – Where general models fall short on spatial
00:16:44 – DuckDB and opinionated workflows
00:18:14 – Data quality as the whole game
00:20:58 – The category problem: supermarket vs. grocery store
00:27:53 – Embeddings and conceptual space, with a 3D walkthrough
00:38:22 – A vision for agentic AI in geospatial
00:43:37 – The end of the scarcity of intelligence
00:47:25 – Where to find Ryan and Dataplor
📰 Daily modern GIS insights:
https://forrest.nycCONNECT WITH ME
📸 Instagram:
https://www.instagram.com/matt_forrest/💼 LinkedIn:
https://www.linkedin.com/in/mbforr/📧 Newsletter:
https://forrest.nyc🌐 Website:
https://forrest.nyc