Vybr 🏠

College roommate matching powered by OpenAI & Qloo Taste AI

Inspiration

Moving to campus can be stressful—finding a roommate who shares your schedule, study habits, music taste, and lifestyle is critical for a happy first year. We wanted to build a social app that goes beyond simple swipes or static questionnaires and uses AI to truly understand your “vibe” and match you with compatible classmates in real time.

What it does

Edu-email OTP signup
- Validates only .edu addresses, sends a 6-digit code via Firebase Functions.
Conversational onboarding
- GPT-powered chat asks about your major, sleep schedule, study habits, cleanliness, social level, music taste, and hobbies.
Taste profile creation
- Feeds your answers into Qloo’s Taste AI™ to build a dynamic “vibe vector.”
AuraMatch recommendations
- Retrieves other campus profiles, ranks them by compatibility score, and surfaces swipeable match-cards with personalized titles and bullet-point reasons.
Instant connect
- Tap “Connect! ” to kick off a chat with your new roommate match.

How we built it

Front-end: React Native with Expo
Auth & Data: Simulated Firebase Functions & Firestore maps for OTP, user profiles, and matches
AI:
- OpenAI GPT-4-Mini for conversational onboarding and personality parsing
- Qloo Taste AI for compatibility scoring and taste profiling
Devops: Hosted Cloud Functions for OTP, Firestore Emulator for local testing
Design: Custom animated chat interface, swipe-card UI for matches

Challenges we ran into

Token latency: waiting on OTP functions vs. smooth UX required loading indicators and delayed navigation.
Intent parsing: NLP sometimes mis-categorized free-form answers—solved via quick-reply buttons and regex fallbacks.

Accomplishments that we’re proud of

Built a full end-to-end demo in under two weeks, from OTP signup through AI-powered matching.
Designed a natural chat UI that feels like texting a friend, not answering a boring form.
Created a clean, modular codebase with easily swap-in OpenAI and Qloo API calls.

What we learned

OTP latency vs. flow: Keeping the UI responsive while waiting on our simulated OTP function required loading spinners and smooth hand-offs between screens.
Parsing open answers: Free-form text sometimes tripped up our simple keyword logic—quick-reply buttons and regex fallbacks helped guide users.
Solo scope: As the only developer, balancing speed of iteration with maintainable code was a constant juggle.

What’s next for Vybr

Real Qloo integration: swap in live API keys for true taste profiling.
Social sharing: let users tweet their “AuraMatch” and drive campus virality.
Profile pictures & socials: allow photo upload & link Instagram/TikTok for richer cards.
Live chat & scheduling: integrate messaging and calendar invites so roommates can plan meetups instantly.