ReBin

Inspiration

Recycling has always been more confusing than it should be. Every city, county, and even building seems to have different rules, and most people genuinely want to recycle — they just don’t know how. Our team kept running into that same frustration, and we thought: what if your recycling bin could teach you back?

That spark led to ReBin Pro — an AI-powered waste sorting application that helps users make environmentally conscious decisions in real time. We wanted to combine modern AI, edge computing, and gamification to make recycling intuitive, accurate, and rewarding.

What it does

ReBin Pro uses AI, geolocation, and real-time data to help users recycle smarter, wherever they are.

Here’s how it works:

📸 AI-Powered Detection: Snap or upload a photo — our model instantly identifies the item using YOLOv8.
📍 Local Policy Integration: It checks your ZIP code and shows exactly how your area wants that item disposed.
⚡ Real-Time Edge Processing: Cloudflare Workers handle image analysis for ultra-fast results and scalability.
🌱 Environmental Tracking: Users see their impact — CO₂ savings, waste diverted, and recycling streaks.
🏆 Gamified Sustainability: Leaderboards, challenges, and achievements make sustainability fun and social.
💬 Community + Social: ReBin connects users through shared challenges, badges, and eco-impact leaderboards.

Everything runs smoothly as a Progressive Web App (PWA) with offline support, push notifications, and a fully responsive design.

How we built it

ReBin Pro is powered by a hybrid AI and edge architecture that keeps it fast, scalable, and intelligent.

Frontend:

Built in React 18 + TypeScript using TailwindCSS for a clean, eco-tech UI.
React Query and Zustand manage data and app state efficiently.
Designed as a Progressive Web App (PWA) for offline support and mobile installation.

Backend:

Developed with FastAPI (Python) for high-performance async APIs.
Supabase handles authentication, real-time updates, and PostgreSQL storage.
Integrated Gemini for reasoning, explanations, and eco-tips.
Computer vision powered by YOLOv8 for waste material recognition.

Edge Layer:

Cloudflare Workers optimize and process images close to the user for instant feedback.
KV storage for caching local recycling policies.
Event tracking, performance monitoring, and rate limiting built directly into the edge layer.

Together, these systems deliver a fast, reliable, and scalable app that feels smooth — even when processing AI predictions in real time.

Challenges we ran into

We ran into a fair share of roadblocks along the way:

🧩 Git merge chaos: With multiple contributors pushing updates quickly, we hit tons of merge conflicts and broken branches. Keeping everything synced across frontend, backend, and model layers took serious coordination.
🤖 Model accuracy & selection: We tested several AI models before landing on YOLOv8. Many struggled with misclassifications or latency, so we fine-tuned and optimized inference pipelines for our use cases.
⏱️ Time constraints: Integrating Gemini reasoning, YOLO detection, Supabase real-time features, and Cloudflare edge processing in a short timeframe was intense. Making sure everything ran smoothly for real users was a race to the finish line.

Each obstacle taught us something new about teamwork, system design, and balancing innovation with practicality.

Accomplishments that we're proud of

Built a fully functional AI-powered waste sorting app with real-time local policy integration.
Deployed an edge-accelerated image pipeline using Cloudflare Workers for sub-second feedback.
Created a gamified community platform that makes sustainability engaging.
Delivered a clean, accessible PWA interface that runs offline and across all devices.
Successfully merged AI, environmental impact tracking, and social engagement into one cohesive experience.

What we learned

AI is powerful — but context matters. It’s not enough for a model to “see”; it has to understand what to do with that insight.
Edge computing changes everything. Processing closer to users through Cloudflare Workers made our app noticeably faster and more scalable.
Version control is teamwork. Clear Git workflows save hours (and sanity).
Design drives adoption. A simple, beautiful interface makes sustainability approachable.
Community multiplies impact. People are far more motivated when they can see their collective results.

What's next for ReBin

We’re just getting started — our roadmap is ambitious:

🧠 Improve model accuracy with multimodal AI (text + vision fusion using Gemini).
🗺️ Expand ZIP-code and international recycling policy coverage.
📱 Launch dedicated mobile apps with AR scanning and real-time push insights.
🌍 Partner with cities, universities, and recycling programs to provide verified data.
💚 Build a global sustainability network — connecting users, governments, and organizations around shared impact metrics.

Our goal is simple: make sustainable action effortless — one toss, one scan, one community at a time.