Inspiration

We wanted to tackle one of the root blockers to solving housing, zoning, and climate challenges in San Francisco: broken, fragmented land data. Existing tools like SF PIM work for parcel lookup but not for analysis or scale. We envisioned a system where anyone — from urban planners to community advocates — could explore the city’s land in real time and ask meaningful, city-wide questions.

What it does

Seekly unifies parcel data from official sources, cleans data, and exposes it through a public API. It includes an interactive map for exploring parcels and a natural language AI assistant that lets users ask questions like “Show me vacant lots near BART” or “Which parcels in SoMa allow mixed use over 5 stories?” Seekly makes complex land data accessible, searchable, and programmable.

How we built it

Aggregated parcels API from official sources .

Cleaned and standardized the data into a consistent schema

Built a backend API to serve parcel queries

Developed a React-based map UI for interactive exploration

Integrated an LLM-powered AI assistant that translates user queries into API calls and displays responses as maps or tables

Challenges we ran into

Data inconsistencies across agencies required custom cleaning and merging logic

Mapping historical zoning codes to their modern equivalents was complex

Translating vague user questions into precise spatial queries pushed the limits of the AI assistant

Designing an experience that worked for both planners and casual users was a constant balancing act

Accomplishments that we're proud of

Built a working platform that unifies and serves real parcels

Designed a natural language interface that works live with our API

Created a clean, scalable backend architecture ready to handle the city’s 300,000+ parcels

Delivered a real-time demo where users can ask spatial questions and see instant results

What we learned

Public data is powerful, but messy — cleaning is often more important than collecting

Planners and technologists think in very different ways, and designing for both requires empathy and iteration

Natural language interfaces can unlock access to data that was previously hidden behind maps and code

What's next for Seekly

Scale to all 300,000+ parcels in San Francisco

Add more data layers: mobility, equity, climate risk, permits, and infrastructure

Improve AI assistant’s reasoning with structured query translation and better context memory

Open the API to civic developers, nonprofits, and planning departments

Built With

Share this project:

Updates