Case Study · AI Product
Smart Home Digest
AI · Scoring · Telegram
Homebuyers get 200+ listing alerts a day. Most are noise. I built an AI that reads a buyer's real priorities and delivers the 3–5 homes that actually matter — with the reasoning visible.
The Problem
Zillow and Redfin are built to surface everything. That's great for the platforms — more alerts means more engagement — but it's the wrong product for a buyer who's already overwhelmed. The signal is buried in the noise. Real estate agents solve this with judgment, but agent judgment isn't available at 7am, isn't personalised to your exact criteria, and isn't scalable.
Users & What I Learned
The homebuyer
Wants to open one message in the morning and know exactly which 3–5 homes to look at. Not a ranked list of 50. Not a spreadsheet. A confident recommendation they can act on.
Critical insight: trust is a product feature
Test buyers override a black-box ranking. They need to see why a home scored 85 vs 45 — which factors drove it, which deal-breaker penalised it. Without transparency, the system gets ignored.
Decisions I Made and Why
Must-haves at 25%
User research showed buyers stretch on budget before they compromise on a must-have. Weighting must-haves above price reflects how buyers actually make decisions.
Telegram, not email
Email digests get archived unread. Telegram gets opened. Channel choice is a product decision — the best ranking algorithm fails if nobody sees the output.
Deal-breaker penalties
Hard exclusions remove homes buyers might want to see with caveats. A −5% penalty per deal-breaker keeps them visible while clearly signalling the tradeoff.
Processing Pipeline
Buyer Profile
Form input captures priorities, must-haves, deal-breakers, and budget.
Fetch Listings
Redfin scraper collects the latest homes matching base criteria.
Normalize
Standardises listing data into a consistent JSON schema for scoring.
Score & Rank
Weighted algorithm scores every home across 7 criteria.
Deliver Digest
Telegram bot sends the top 3–5 homes with reasoning.
Pipeline: Buyer profile → listing fetch → normalisation → weighted scoring → curated Telegram digest.