Inspiration
What it does
NYC grades every restaurant. Nobody grades landlords. Paste a Bronx rental listing and we fact-check it against city records.
What it does: Renters sign leases blind. Before You Sign grades every Bronx apartment building A–F, per apartment, using 167,819 open hazardous HPD violations and 40,754 marshal evictions.
Paste a listing ("pest-free, lead-safe, great heat") and it checks each claim against the records, then marks it TRUE, FALSE or CAN'T CHECK with a dry one-liner. It suggests questions to ask the landlord. It finds better-graded buildings a few blocks away. It answers in English, Spanish or Arabic. It links to a visual report card with a map, a household lens (kids, asthma, older adults) and the nearby alternatives.
How we built it:
Elastic: Serverless Elasticsearch with two NYC Open Data indices, plus a Kibana Agent Builder agent with 8 ES|QL tools. The grading happens in Elastic, not the LLM. Grades are computed in ES|QL across both indices. A geospatial query uses ST_DISTANCE plus INLINE STATS to find better buildings within 800 m. Claim verdicts are also computed in ES|QL, so the model can't invent a verdict. Mistral: Mistral Large through the Elastic Inference API reasons over the tools, and writes the plain-language explanations, the deadpan humor and the multilingual answers.
Built With
- elasticsearch
- es|ql
- kibana-agent-builder
- mistral-ai
- nyc
- open
Log in or sign up for Devpost to join the conversation.