Berkeley-specific real estate analytics for people who want the truth.
HomeBuyer is a side project that started because one person wanted to understand Berkeley real estate and got slightly carried away. It combines ML price predictions, an AI-powered property analyst, neighborhood analytics, zoning intelligence, and investment scenario modeling into a single tool that tells you what a house is actually worth — not what someone trying to earn a commission wants you to believe.
This is not a real estate brokerage. This is not financial advice. This is a developer with too much free time and strong opinions about housing prices.
- An ML price prediction model trained on 5 years of Berkeley sales data with ~$95K median absolute error
- An AI property analyst (Faketor) powered by Claude with 18 tools for comps, zoning, permits, rental estimates, investment analysis, and more
- A neighborhood analytics dashboard with median prices, trends, days on market, and sale-to-list ratios for every Berkeley neighborhood
- A zoning and development tool covering ADU feasibility, SB 9 lot splits, Middle Housing rules, and setback requirements
- An affordability calculator that tells you what you can actually afford (spoiler: it's less than you think)
- An investment prospectus generator that models flip, rent, and hold scenarios with actual math
- A licensed real estate brokerage or advisory service
- A substitute for professional real estate, legal, or financial advice
- A guarantee of anything — the model is good but it's not omniscient
- A way to make money (we certainly haven't)
For the full legal version, see the Terms and Conditions (accessible from the app's login and settings pages).
HomeBuyer pulls from a variety of public and third-party data sources:
Public (free):
- Redfin — Property sales history and market metrics via their GIS-CSV API and Data Center
- FRED — 30-year and 15-year mortgage rates, plus economic indicators (NASDAQ, 10-year Treasury, CPI, consumer sentiment, unemployment)
- U.S. Census Bureau — American Community Survey median income by zip code
- Berkeley Open Data — BESO energy benchmarking and parcel boundaries
- Berkeley Municipal Code — Zoning regulations (Title 23)
- City of Berkeley — Transfer tax rates, permitting info, Middle Housing policy
- Accela Citizen Access — Building permit records
- Berkeley Rent Board — Rent control information
- FHFA — Conforming loan limits
Third-party (paid, required for full functionality):
- RentCast — Property details, sale history enrichment, and rental estimates
- Anthropic Claude API — Powers the Faketor AI agent
Rebuilding note: If you're looking to rebuild or fork this project, you'll need a paid property data service like RentCast or ATTOM to get the enriched property data that makes the ML model work well. The free public data sources alone won't give you enough features for accurate predictions.
For a detailed breakdown of the API endpoints, Faketor AI tools, CLI data pipeline, and ML model architecture, see SERVICES.md.
# Clone the repo
git clone https://github.com/yordsel/HomeBuyer.git
cd HomeBuyer
# Ask Claude to get things up and running
# (it knows the project structure and will handle dependencies,
# database setup, and dev server configuration)That's it. Claude knows this codebase. Tell it what you want to do and it'll figure out the rest — install Python and Node dependencies, initialize the database, seed the data, and start the dev servers.
You'll need:
- Python 3.12+
- Node.js 18+
- API keys for any third-party services you want to use (see SERVICES.md)
| Layer | Technology |
|---|---|
| Backend | Python, FastAPI, Uvicorn |
| Frontend | TypeScript, React, Vite, Tailwind CSS |
| ML | scikit-learn (HistGradientBoosting), SHAP |
| Maps | Leaflet, react-leaflet |
| AI | Anthropic Claude (via Faketor agent) |
| Database | PostgreSQL (production), SQLite (development) |
| Resend | |
| Auth | JWT + bcrypt, Google OAuth |
| Scraping | Playwright, BeautifulSoup4 |
| Deployment | Render |
This is a side project. Use it, learn from it, fork it. Just don't pretend it's financial advice.