An AI real estate agent - smart property search and decision intelligence that goes beyond a listings database. Built solo, full-stack: frontend, backend, and infrastructure.
Roosthaven is an AI real estate agent - a platform that combines property search with decision intelligence to help buyers make smarter, more informed choices. Instead of being just another listings database, it uses AI to evaluate properties and surface the ones that genuinely fit.
I built Roosthaven solo and full-stack - frontend, backend, server, and everything in between.
The front end is Next.js, scaffolded quickly with v0 (an AI UI tool) and then refined into a production interface. The backend is Node.js, with NeonDB (serverless Postgres) as the data layer and the AI logic that powers search and decision intelligence.
Roosthaven runs on AWS with a GitHub-based CI/CD pipeline - automated builds and deploys so I can iterate solo at speed without manual release steps.
For most people, buying a home is the biggest decision they’ll make with the least data. Roosthaven flips that - putting search and intelligence on the buyer’s side.
I built every layer myself - the Next.js front end, the Node.js backend and APIs, the database, the AI logic, and the AWS infrastructure. Owning the whole stack meant making pragmatic calls about where to spend effort and where to lean on tools like v0 and NeonDB.
The hard part of an 'AI real estate agent' is turning model output into decisions a buyer can trust. The work was in grounding the AI in real property data and shaping it into intelligence that actually helps people choose.
Real estate decisions are high-stakes and data-poor for most buyers. Roosthaven puts intelligence on their side.