Aug 2025 — Present
Lead Developer · CoLive — Roommate Finder App
CoLive started as a Lovable-generated prototype to quickly validate the roommate-matching concept. I took it from there and rebuilt it into a real production application — redesigning the database schema for graph-based compatibility matching, replacing prototype auth with proper OAuth 2.0, building real-time chat infrastructure, and deploying it independently across web, mobile, and API layers.
Overview
CoLive is a roommate-matching platform, and its origin story is the most honest thing about it: it didn't start as a from-scratch engineering project. It started as a Lovable-generated prototype — a fast way to validate whether the roommate-matching concept actually held up before investing real engineering time into it. Once the idea proved out, the real work began: taking a vibe-coded prototype and rebuilding it into something production-grade.
Tech Stack
- React Native
- Expo
- Node.js
- Hono
- Drizzle
- Supabase (PostgreSQL)
- pgGraph
- OAuth 2.0
- Cloudflare Workers
Key Contributions
- Redesigned the database schema to support graph-based roommate compatibility matching using Supabase (PostgreSQL) and pgGraph for real-time relationship queries — a fundamentally different data model than a typical prototype generates.
- Replaced prototype-level auth with proper OAuth 2.0, closing the gap between looking like it works and being secure enough for real users to trust with their data.
- Built real-time chat messaging infrastructure with Node.js, Hono, and Drizzle, enabling live conversation between matched roommates.
- Deployed and configured cloud infrastructure on Cloudflare Workers independently across the React Native (Expo) mobile app, web app, and API layer.
- Designed the user onboarding and matching workflow from first principles, mapping out the full experience before writing implementation code.
The Real Engineering Challenge
Anyone can generate a working prototype in an afternoon now. The harder — and more valuable — skill is knowing what to keep and what to rebuild once that prototype proves the idea works. A vibe-coded app is good at exactly one thing: showing you a working UI fast enough to react to. It is not good at security, data integrity, or scaling past a handful of test users. The judgment call — and the actual engineering work — was in identifying precisely which parts of the prototype could stay and which needed to be torn out and rebuilt before real users could depend on it.
Product Preview
Before → After
Before
Lovable-generated prototype — functional enough to validate the concept, but built on prototype-grade auth, a simple relational structure with no compatibility-matching logic, and no real-time infrastructure.
After
Production application with graph-based matching, OAuth 2.0 authentication, real-time chat, and independently managed cloud infrastructure across web, mobile, and API.
Why It Matters
This isn't a “look what AI can build” story — it's the opposite. It's proof that the valuable skill going forward isn't refusing to use fast prototyping tools, it's knowing exactly where their output ends and real engineering has to begin. That judgment — not the initial prototype — is what makes CoLive a real product rather than a demo.