nextplate

The Waste Is Invisible. The Hunger Isn't.

Comment

Inspiration

Inspiration

Every night, perfectly edible food is thrown away while millions go hungry.

🌍 1.3 billion tons of food wasted annually πŸ› 828 million people still facing hunger

The problem isn’t production β€” it’s coordination.

NextPlate was inspired by a simple but powerful question:

What if surplus food could move as fast as information?

We wanted to build a system where:

Restaurants don’t waste food NGOs don’t struggle to find it People can actively contribute to impact 🎬 The Story (Real-World Flow)

It’s 7 PM on a Tuesday.

A restaurant has 30 extra meals.

They open NextPlate Click a photo πŸ“Έ AI fills details automatically Listing goes live in 40 seconds

Nearby NGO gets notified β†’ accepts pickup Volunteer arrives β†’ scans QR β†’ AI verifies

βœ” 30 meals rescued βœ” ~12kg COβ‚‚ saved βœ” 30 people fed

Zero waste. Real impact. Instantly.

🧠 What I Learned

Building NextPlate wasn’t just coding β€” it was solving a real-world system problem.

πŸš€ Technical Learnings Designing multi-role systems (Restaurant, NGO, User, Admin) Building real-time systems with Socket.IO Integrating AI pipelines with fallback strategies Handling rate limits + retries (429 handling) Implementing geo-based queries with MongoDB Creating scalable REST APIs (60+ endpoints) πŸ€– AI Learnings Multi-key architecture for parallel AI workloads Vision AI (photo β†’ structured data) AI fallback chain: Primary Model β†’ Lite Model β†’ Deterministic Output Primary Modelβ†’Lite Modelβ†’Deterministic Output Designing AI that is useful even when it fails πŸ—οΈ How I Built It πŸ–₯️ System Architecture Frontend (Next.js 16 + React 19) ↓ Backend (Node.js + Express) ↓ MongoDB + Gemini AI + Firebase πŸ”§ Key Components ⚑ Frontend: Next.js + Tailwind + Framer Motion πŸ”Œ Backend: Express.js (20 routes, 18 controllers) 🧠 AI Engine: Gemini 2.0 (11 features) 🌍 Database: MongoDB Atlas (Geo queries) πŸ”” Realtime: Socket.IO + Firebase FCM πŸ’³ Payments: Razorpay πŸ“© Communication: SendGrid + Twilio πŸ€– AI Engine (Core Innovation)

NextPlate runs on a multi-key AI architecture:

πŸ”‘ Chat Key β†’ Conversations πŸ”‘ Analytics Key β†’ Predictions πŸ”‘ Vision Key β†’ Image Processing βš™οΈ Smart Retry + Fallback

If AI fails:

Retry Delay

2 𝑛 seconds ( 2 𝑠 , 4 𝑠 , 8 𝑠 ) Retry Delay=2 n seconds(2s,4s,8s)

Then fallback:

Flash β†’ Lite Lite β†’ Deterministic logic

This ensures 100% system reliability

🌿 Impact Calculation (Carbon Saving)

We implemented WRAP methodology:

COβ‚‚e Saved

Food Weight (kg) Γ— Carbon Factor COβ‚‚e Saved=Food Weight (kg)Γ—Carbon Factor

Example:

30 meals β‰ˆ 12 kg COβ‚‚ saved 30 mealsβ‰ˆ12 kg COβ‚‚ saved

Also tracked:

πŸ’§ Water saved (~1000L per kg food) 🍽 Meals rescued 🌍 Community impact ⚑ Key Features Built πŸ“Έ Stitch Vision β†’ Photo β†’ auto listing 🧠 AI Rescue Engine β†’ NGO decision making πŸ›’ Surplus Marketplace β†’ Discounted food πŸ‘» Ghost Meals β†’ Sponsor food for others πŸ“Š Impact Dashboard β†’ Carbon + meals tracking πŸ” QR Verification β†’ Secure pickup πŸ—ΊοΈ Live Map β†’ Real-time food availability πŸ› οΈ CLI Tool β†’ Monitor global food grid βš”οΈ Challenges Faced

  1. πŸ”₯ AI Rate Limits (429 Errors) Solved with: Exponential backoff Multi-key distribution Model fallback
  2. 🧠 Making AI Reliable

AI can fail β€” but products cannot.

Solution:

Built deterministic fallbacks Ensured graceful degradation

  1. 🌍 Real-Time Coordination

Handling:

Live orders Notifications NGO pickups

Solution:

Socket.IO event system City-based subscriptions

  1. πŸ“ Supply-Demand Matching

Problem:

NGOs nearby couldn’t see food

Solution:

Geo-indexing (MongoDB) Live map + push notifications πŸš€ What Makes NextPlate Unique ⚑ 40-second food listing πŸ€– 11 AI-powered features 🌍 Real-time redistribution network πŸ“Š Measurable environmental impact πŸ” Trust layer with AI verification πŸ‘» Community-powered food sponsorship 🧭 Future Vision πŸ“± React Native mobile app β›“ Blockchain transparency for donations πŸ› Government food safety API integration πŸ“ˆ ML-based demand prediction 🌐 Global expansion of food grid 🏁 Final Thought

NextPlate is not just an app.

It’s a living network where:

Restaurants reduce waste NGOs get resources instantly People contribute to real impact

Every meal rescued is a life impacted.

❀️ Built With Purpose

NextPlate β€” Turning Surplus into Hope.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for nextplate

Built With

Share this project:

Updates