EduConnect

Inspiration

Picture this: It’s 11:47 PM. A calculus assignment is due tomorrow. You’ve reread the prompt six times. Every friend is asleep. Every campus tutoring center is closed. ChatGPT is not helpful enough to guide you through every step. The tutors you can access are either randomly paired, just like on platforms such as NetTutor, or come with a price tag. Most of the time, they’re inflexible, and you have to schedule meetings days in advance. The panic sets in, and you realize: there’s no easy way to get the right person to help you without wasting your time, exactly when you need it. That’s the moment we realized: students need an efficient way to get the right help, something no other platform provides.

What it does

• Algorithmic AI Matching: Instantly finds the best tutor for you based on your preferences, learning style, and class needs.
• Chatbot AI with Video Explanations: Provides personalized step-by-step explanations and guidance tailored to how you learn.
• Real-Time Chat & Whiteboard: Communicate and solve problems with your tutor in real time.
• Easy Rematch Option: Quickly find a new tutor if the first match isn’t ideal.
• Group Studies: Collaborate with peers and tutors in study sessions.
• Personalized Matches: Tutors are chosen specifically to fit your learning style and goals.
• Free to Use: Access all features without paying.

How we built it

Frontend

• Next.js 15, React 19, TypeScript
• Custom swipeable UI with animated card-stack transitions
• Tailwind CSS + custom gradients/glassmorphism
• Responsive layout optimized for both web and mobile -HTML5 Canvas API for collaborative drawing

Backend

• Firebase Authentication (Email/Password + Google OAuth)
• Firestore Database for profiles, swipes, matches, chats, canvas data
• Firebase Storage for profile images with client-side compression
• Real-time listeners (onSnapshot) for chat + canvas synchronization
• Next.js API Routes for AI tutor and serverless logic

Machine Learning / AI

• Custom weighted scoring model (interests, school, schedule, age, personality)
• OpenAI GPT-3.5-turbo for AI tutoring responses
• Client-side image processing with browser-image-compression
• Intelligent match recommendations based on multi-factor scoring

Architecture Summary

• User actions → Firestore (profiles, swipes, chats)
• Matching Algorithm → calculates weighted compatibility scores
• AI Tutor API Route → handles OpenAI requests
• Frontend → displays recommended matches, swipe UI, chat, and canvas

Challenges we ran into

• Building a clean, intuitive swipe UI without external libraries
• Syncing real-time chat with live match updates
• Managing Firestore reads/write limits efficiently
• Designing a fair, bias-free recommendation algorithm
• Tuning the ML model with limited real-world data
• UX design: balancing simplicity with powerful features
• Real-time location-based matching without compromising privacy

Accomplishments that we're proud of

• Built a fully functional Tinder-style swipe interface
• Created a complete two-stage AI matching pipeline
• Delivered real-time match + chat functionality
• Designed a clean, student-friendly UX
• Ensured everything works on mobile and desktop
• Built a system that can genuinely reduce student stress
• Started with an idea → turned it into a polished demo in 48 hours

What we learned

• How to integrate Firebase + ML smoothly
• How to design intuitive user flows for stressed students
• How to handle swiping logic, card animations, and match states
• How to use embeddings for semantic matching
• How to build explainable AI recommendations
• How powerful ML can be in solving everyday student problems

What's next for EduConnect

• Adding course-specific study rooms
• Advanced scheduling & calendar sync
• Verifying helpers via school email
• Better safety tools + in-app trust system
• Group study matching
• Multi-modal models for understanding images of handwritten notes

Built With

• css
• geminiapikey
• googleantigravity
• html
• javascript
• json
• localdatabase
• react
• typescript