Discover DMV
Product · Engineering
A field guide to the DC area's speakeasies — found, verified, and mapped for a great night out.
Behind the unmarked door
The DMV — DC, Maryland, and Virginia — hides some wonderful bars behind unmarked doors and quiet storefronts. Discover DMV is a deliberately niche app for finding them: a curated catalog of the area's speakeasies.
Under the hood it runs on a data pipeline I built. An AI extraction system reads local-business websites and pulls structured information — hours, locations, contact details — and then I verify every field by hand against its source before it enters the dataset. Human-in-the-loop, so the data can be trusted.
The result is a geocoded dataset of the area's restaurants and bars, with speakeasies as the first curated collection.
Field research
In the app
How it's built
- TypeScript on Cloudflare Workers — a low-cost, serverless pipeline with a D1 (SQLite) database.
- AI extraction — structured data pulled from local-business websites into a consistent schema.
- Human-in-the-loop verification — every extracted field is checked against its source before it counts.
- Geocoded dataset — every spot placed on the map, ready for discovery.
What's next
Currently building: the discovery experience over the verified dataset — browse the catalog, see the map, plan the night.