Skip to content

About

MustangMatch — AI-matched friend finder for Cal Poly students, built for Codebox Hacks

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

MustangMatch 🐎

An AI-matched friend-finder for Cal Poly SLO students. Built for Cal Poly Codebox Hacks (independent submission track — Track 3/4).

The problem

Cal Poly has 22,000 students spread across dozens of majors, clubs, and residential communities. Meeting people outside your immediate dorm hall or major is mostly luck. Existing social apps (Instagram, GroupMe) assume you already know who to add — they don't help you find anyone.

What it does

  1. You sign up and tell it your year, major, hometown, and interests in your own words.
  2. Claude reads your interests and normalizes them into tags, then — for every other student in the pool — generates a genuine compatibility score and a one-line reason you two might click, going beyond exact keyword matching (e.g. it treats "hiking" and "backpacking" as related).
  3. You get a Tinder-style swipe deck of Cal Poly students ranked by that score, with the shared interests highlighted.
  4. Mutual likes unlock a match — and whatever contact info the other person chose to share.

Tech stack

  • Next.js 16 (App Router, TypeScript) — full-stack React app, deployed on Vercel.
  • Supabase (Postgres + Auth) — email/password auth and the students/swipes/matches data.
  • Anthropic Claude API — interest normalization + pairwise compatibility scoring and blurb generation, called server-side from Next.js Route Handlers and cached per-pair so repeat views don't re-call the model.
  • Tailwind CSS v4 + Framer Motion — Cal Poly green/gold themed UI and the swipe card animation.

All data reads/writes go through server-side Route Handlers using Supabase's service role key, after verifying the caller's session — the browser never talks to the database directly.

Local setup

  1. npm install
  2. Copy .env.example to .env.local and fill in your Supabase + Anthropic keys.
  3. In the Supabase SQL editor, run supabase/schema.sql once to create the tables.
  4. (Optional but recommended for demoing) seed ~36 fake Cal Poly profiles so the swipe deck isn't empty: node scripts/seed.mjs
  5. In Supabase → Authentication → Providers → Email, turn off "Confirm email" so sign-up logs you in immediately (fine for a demo; turn it back on for anything real).
  6. npm run dev

Deploying

Push to GitHub, import the repo in Vercel, add the same environment variables from .env.local in the Vercel project settings, and deploy.

Scope notes (what's intentionally left out)

Built in under a day for the hackathon deadline, so a few things were deliberately cut to ship something real instead of something half-built:

  • No photo upload — profiles use a generated initials avatar instead.
  • No in-app messaging — a match reveals whatever contact info (e.g. Instagram) the person chose to share, rather than building a full chat system.
  • Demo profiles are seeded fictional students so the matching experience is fully testable without needing 30+ real signups first.

About

MustangMatch — AI-matched friend finder for Cal Poly students, built for Codebox Hacks

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages