Overview
CompLens turns the slowest part of underwriting a multifamily deal (pulling comps by hand) into a single search. You type an address, and it returns a structured comp memo: nearby multifamily properties ranked by rent per square foot, demographic context from the Census, and everything plotted on a map.
I'm building it solo, full stack, end to end. The interesting part is the data: instead of leaning on listing aggregators, a two-model Claude pipeline sources asking rents from official property leasing sites first, so the comps reflect what landlords are actually quoting.
The problem
Pulling rent comps is manual and scattered: aggregator data is stale or padded with fees, official leasing sites all present pricing differently, and analysts end up stitching together screenshots and spreadsheets for every address they underwrite.
Approach
- Built the product solo, full stack: React + Vite frontend deployed on Vercel, Supabase for auth and Postgres.
- Integrated Google Maps APIs for geocoding, nearby-property discovery, and the map view.
- Designed a two-model Claude pipeline that prioritizes official property leasing sites over aggregators when sourcing asking rents.
- Layered in Census demographics so each memo carries neighborhood context, not just rent figures.
Outcomes
- One search produces a structured comp memo that used to take an afternoon of manual pulls.
- Comps ranked by rent per square foot with source-aware provenance for each figure.
- Live now, with coverage and memo depth still growing.
Stack & tools
React, Vite, Vercel, Supabase (auth + Postgres), Google Maps APIs, Claude (two-model pipeline).