Problem

I set out to help people find neighborhoods that match their lifestyle, not just their budget. The challenge was translating user preferences into a reliable, scalable livability score across over 400 metro locations in the NYC, SF, and LA areas. Consumers needed a familiar, dashboard-friendly view that reflects what matters to them—walkability, schools, transit, nightlife, and green space—while ensuring the data remained comparable across districts with distinct boundaries.

I also faced data governance and accuracy concerns. School ratings, for example, required a method that accounts for access differences and avoids drawing broad conclusions from limited samples. Geographic boundaries needed careful handling to prevent scores from bleeding across unrelated zones. A fast, maintainable pipeline was essential to deliver timely results to the frontend without sacrificing fidelity or scalability.

Solution

I built a scoring API (Python/FastAPI) deployed on Railway and a frontend on Vercel at trovamo.co to deliver dynamic, per-location scores across 13 pillars. Each pillar has a custom methodology—for instance, school scores use the lower-two-thirds mean of ratings to reflect typical family access, plus an access bonus for elite schools. This approach preserves nuance while staying comparable across neighborhoods.

Geographic routing logic anchored NYC to Community School Districts (32 districts via centroid lookup) and LA to ZIP-based pools to prevent cross-border score blending. I developed a catalog pipeline that scores 290 metro locations and produces merged JSONL catalogs for fast frontend delivery. The stack—Python, FastAPI, Railway, Vercel, and geocoding APIs—supports iterative refinements and scalable deployments.

Trovamo - HomeFit