Connectify

Find your tribe, share your vibe! Connect with fellow students who truly get you!

Comment

Inspiration

Connectify was inspired by the growing loneliness epidemic among college students in the US. Many students struggle to find meaningful connections on campus, especially those who are introverted, new to the college, or have niche interests. We noticed that traditional social platforms often prioritize quantity over quality and don't focus on helping students connect with peers they can actually meet in person.

The idea came from recognizing that the best friendships often start with a single shared thought, feeling, or passion and that these connections are most meaningful when they happen with people from your own college community. We wanted to create a platform that uses the power of AI to match students based on the depth of their interests rather than superficial metrics, while ensuring they can actually meet and build relationships on campus. This college-specific approach helps combat loneliness by fostering authentic, local connections that can grow into lasting friendships.

What it does

Connectify is an AI-powered social connection platform designed specifically for college students in the US who feel lonely or struggle to connect with peers. It helps students find friends from their own college based on shared emotions. Here's what it offers:

  1. College-Specific Interest Matching: Users select their college from a dropdown of 50+ US colleges and share their thoughts and feelings. The platform converts these into vector embeddings using OpenAI's API and finds students from the same college with similar interests using cosine similarity. This ensures students connect with peers they can actually meet and build relationships with on campus.

  2. Smart Chat System:

    • AI-generated ice breaker messages from CONNECTIFY that introduce users
    • Personalized questions related to shared interests
    • Real-time conversation with context awareness

  3. Scheduling & Meetups:

    • Users can schedule meets with date, time, and location
    • RSVP system with celebration animations

  4. Memory Wall:

    • Beautiful cork board-style display of positive reflections
    • Shows memories from people who have actually met
    • LED string lights connecting memories for a cozy, nostalgic feel
    • Polaroid-style cards with pushpins

  5. Visual Design:

    • Modern, colorful gradient UI
    • Dancing friends animation in loader
    • Responsive design for all devices

How we built it

Backend (FastAPI + Python)

  • Vector Embeddings: Used OpenAI's text-embedding-3-small model to convert user queries into 1536-dimensional vectors
  • Database: Supabase PostgreSQL for storing query embeddings with vector similarity search and college filtering
  • College Filtering: Database schema includes college column with index for efficient filtering. Similarity search automatically filters matches to only show students from the same college
  • AI Integration: OpenAI GPT-3.5-turbo for:
    • Generating ice breaker messages
    • Generating positive reflections for Memory Wall
  • Similarity Matching: Cosine similarity algorithm to find users with similar interests (50%+ threshold) within the same college

Frontend (React + TypeScript)

  • Framework: React with React Router for navigation
  • UI Components: Custom components built with Tailwind CSS and shadcn/ui
  • State Management: React hooks and localStorage for tracking met friends
  • Animations: CSS animations and transitions for smooth user experience
  • SVG Graphics: Custom SVG illustrations for logo and Memory Wall decorations

Key Technologies

  • Frontend: React, TypeScript, Tailwind CSS, Vite
  • Backend: FastAPI, Python, OpenAI API, Supabase
  • Vector Search: PostgreSQL with pgvector extension (via Supabase)
  • AI Models: OpenAI GPT-3.5-turbo, text-embedding-3-small

Challenges we ran into

  1. Vector Embedding Storage: Initially struggled with storing high-dimensional vectors in Supabase. Solved by using JSONB columns and proper vector normalization.

  2. Similarity Threshold Tuning: Finding the right similarity threshold (50%) required testing to balance between too many irrelevant matches and too few matches, especially when filtering by college. 3.State Management: Tracking which friends have actually "met" required implementing localStorage persistence and ensuring data flows correctly between chat, scheduling, and Memory Wall components.

  3. Seamless AI Integration: Successfully integrated multiple OpenAI endpoints (embeddings, chat completions) to create a cohesive user experience.

  4. Beautiful Memory Wall: Created an aesthetically pleasing Memory Wall with:

    • Realistic cork board texture
    • Organic card placement with hover interactions
    • Dynamic LED string lights connecting memories
    • Polaroid-style design with pushpins

  5. Smart Scheduling System: Built an intuitive scheduling feature with:

    • Keyword detection ("meet") triggering automatic scheduled connects
    • RSVP system with celebration animations
    • Beautiful modal interface

  6. Dancing Friends Loader: Created a unique, branded loading animation using the same SVG friends from the logo, making the wait time more engaging.

  7. Comprehensive User Journey: Designed a complete flow from interest sharing → matching → chatting → scheduling → memory collection.

  8. Responsive Design: Built a fully responsive application that works beautifully on desktop, tablet, and mobile devices.

  9. Demo-Ready Features: Implemented demo data and features that showcase the platform's potential even without real user interactions.

What we learned

  1. Vector Embeddings: Gained deep understanding of how vector embeddings work and how to use them for semantic similarity matching. Learned about cosine similarity, normalization, and embedding dimensions.

  2. AI Prompt Engineering: Discovered the importance of carefully crafted prompts for generating natural, varied, and contextually appropriate AI responses. Learned to balance creativity with consistency.

  3. SVG Animation: Mastered creating complex SVG animations and understanding coordinate systems, transforms, and path calculations for dynamic visual effects.

  4. State Persistence: Learned best practices for managing application state across components and persisting user data (like met friends) using localStorage.

  5. User Experience Design: Gained insights into creating intuitive interfaces that guide users through complex workflows (matching → chatting → scheduling → memories).

  6. Full-Stack Integration: Learned to seamlessly connect frontend React components with backend FastAPI endpoints, handling async operations and error states gracefully.

  7. Database Design: Understood how to structure data for vector similarity searches and optimize queries for performance.

What's next for Connectify

  1. Real-Time Messaging: Replace AI agent simulation with actual real-time messaging between users using WebSockets or similar technology.

  2. Enhanced Matching Algorithm:

    • Implement multi-factor matching (interests, location, availability)
    • Add machine learning to improve match quality over time
    • Consider user feedback to refine recommendations

  3. Social Features:

    • User profiles with photos and detailed interests
    • Friend groups and communities
    • Event creation and group meetups
    • Activity feed showing friend connections

  4. Mobile App: Develop native iOS and Android apps for better mobile experience and push notifications.

  5. Advanced Memory Wall:

    • Allow users to add photos to memories
    • Timeline view of connections
    • Share memories with friends
    • Memory search and filtering

  6. Safety & Moderation:

    • Content moderation system
    • User reporting and blocking features
    • Verification system for scheduled meets
    • Safety tips and guidelines

  7. Analytics & Insights:

    • Show users their connection patterns
    • Interest trends and popular topics
    • Success metrics for friendships formed

  8. Integration Features:

    • Calendar integration for scheduling
    • Location-based suggestions
    • Social media profile linking
    • Interest-based group recommendations

  9. Gamification:

    • Friendship score tracking
    • Achievement badges
    • Connection streaks
    • Leaderboards for most active connectors

Connectify has the potential to revolutionize how people form meaningful connections in the digital age, combining the power of AI with human warmth and genuine interest in building authentic friendships.

Built With

Share this project:

Updates