Project goals were to build an AI‐driven real estate analysis platform that models and compares new‐construction scenarios—home styles, sizes, garages, basements, prices and locations—to pinpoint the most profitable builds on each lot.
Project Details
- Built and deployed a Python Selenium web-scraping bot to log into our MLS portal and extract 365-day data on active, pending, and sold homes—capturing styles, garage count, square footage, basement finishing, prices, and geolocations.
- Ingested and harmonized MLS data via pandas and SQLAlchemy into a unified dataset of home attributes (ranch, 2-story, 1.5-story; above-ground size; attached garages; basement specs; sale price; location).
- Implemented a rule-based pro forma engine that calculates build costs, timelines, and sale schedules for customizable development scenarios.
- Developed an XGBoost-powered Automated Valuation Model (AVM) using historical comps to predict future home values and feed profit-and-risk KPIs.
- Modeled scenario KPIs—projected profit, ROI, time-to-sell—across hundreds of lot-and-home-style combinations under varying cost and market assumptions.
- Embedded regulatory constraints (zoning rules and quota limits in the American housing market) into the scenario engine so only compliant build options surface.
- Generated actionable recommendations, ranking lots by their highest-yield home style and configuration, complete with dynamic dashboards and exportable reports.
- Delivered full documentation and reproducible code, enabling seamless hand-off and future enhancements by the client’s in-house team.