Case Study - AI-powered home search that understands you

Overview

The home search experience has been broken for years. People spend hours scrolling through listings, applying endless filters, and still miss homes that would be perfect for them. Homi was built to solve this problem with AI that actually understands what you're looking for.

Instead of traditional search filters, Homi uses conversational AI to learn about your preferences, budget, location requirements, and lifestyle needs. The system analyzes listings across multiple sources and presents homes that genuinely match, not just on paper, but in the ways that matter for how you'll actually live there.

The platform launched with immediate traction: 250+ organic signups validated the problem, broker partnerships provided access to premium inventory, and early revenue proved the business model. The technology stack (Next.js, TypeScript, and tRPC) enabled rapid iteration while maintaining code quality and type safety.

The next phase involves expanding data sources, deepening AI capabilities to understand neighborhood dynamics, and building features for the entire home search journey from discovery to closing.

The first product map

Before Homi had an AI search experience, I mapped the home-hunting job as four steps: collect, organize, compare, and choose. Each step was connected to the pages, data, and user stories needed to support it.

The screenshot below is the 2024 marketing page for that first version. It is now part of Product Atlas, an evidence-based record of how the product has changed and which parts of the original job remain.

What I built