Comparables Explorer Comparables Explorer

Overview

Built a password-gated comparables explorer for a real-estate client that unified 36,027 Guadalajara sale and rental listings into median-priced zone benchmarks, with a below/at/above-market verdict for every property.

Highlights

1. 01

   Engineered a read-only Python ETL (Postgres staging → Parquet → versioned JSON contract) that normalizes currency, separates land from built area and winsorizes prices per segment, so the static explorer runs with no database at runtime.

2. 02

   Designed an adaptive-cohort pricing verdict (N≥8, ±15% bands) computed once in the ETL, verified at 0 contradictions between map pins and detail panels across 19,034 sale listings.

3. 03

   Replicated the product for a second Mexico City market (565 listings from 3 sources, 100% geolocated) and hardened it with XSS escaping, self-hosted map libraries, CSP headers and ARIA-compliant controls.

Stack

• Next.js
• Python
• pandas
• PostgreSQL
• Parquet
• Leaflet
• Leaflet.markercluster
• Winsorization
• Median-Based Benchmarking
• Vercel
• Content Security Policy