By the numbers
What I built
The resolution engine behind Block Party
Block Party, run with Sarah Sachs, lets residents and board members search what their board decided and why. My part is the resolution engine: the pipeline that collects each board's minutes, pulls out every resolution with its text, vote and date, removes duplicates, and tags each one with one of 28 topics.
How it works
Minutes in, a searchable record out
- Scrapers collect each board's posted minutes and agendas, going back to 2002 in Manhattan and 2006 in Brooklyn.
- A language model extracts each resolution's text, vote tally and date; code checks dates and removes duplicate committee and full-board copies.
- A zero-shot classifier tags every resolution with one of 28 topics, from liquor licenses to open streets.
- A weekly automated run adds new minutes, and every expansion goes through a blind audit against the source minutes before launch.
I presented the project at NYC School of Data (March 2026), the NYC OMB Institute (July 2026, with Sarah Sachs) and on the main stage of CivStart's State of GovTech 2026 in Washington, DC (September 2026).
What a commission like this looks like
Any organization with years of records can search them
Community boards, civic groups, councils and agencies keep years of minutes, testimony and decisions that the public cannot search in one place. I build the same kind of archive on your records: collection, extraction, topic tagging and a search interface, with an automated refresh. Scope depends on how many documents you have and what shape they are in.
More case studies
All NYCuriosity Studio services → NYC Council Legislation Trackers → State Capacity Ecosystem →