Inspiration
Food waste and improper waste disposal are two everyday problems that are often treated separately. We wanted to build a single platform that helps people make better decisions about what they throw away and what they can still use.
The idea behind EcoSort AI came from a simple question: What if technology could help us understand our waste before it becomes waste?
Instead of simply telling users to "recycle more," we wanted to create an AI-powered platform that gives people an immediate, practical action. A photo of an item can tell them how it should be disposed of, while leftover food can be transformed into recipes, meal plans, or opportunities to share surplus food.
Our goal is to connect everyday decisions with sustainability and make responsible consumption easier.
What it does
EcoSort AI is an AI-powered waste management and food waste reduction platform designed around SDG 12: Responsible Consumption and Production.
AI Waste Classifier
Users can upload a picture of a waste item and EcoSort AI identifies the appropriate category:
- ♻️ Recycling
- 🌱 Compost
- 🗑️ Landfill
- 🔋 E-waste
- ⚠️ Hazardous waste
The system also provides a confidence score, disposal guidance, and sustainability tips so users understand what to do next.
Digital Fridge
Users can track food and ingredients in a digital fridge, including expiry information and available ingredients.
AI Recipe Matcher
Instead of throwing away ingredients that are approaching their expiry date, users can enter or scan available ingredients and receive recipe suggestions. Our recipe system uses a HuggingFace model with catalog/web-search fallback.
Meal Planner
EcoSort AI can create multi-day meal plans and generate grocery lists based on what users already have, helping reduce unnecessary purchases.
Surplus Sharing
Users can share surplus food with people nearby, creating another pathway for edible food to be used instead of discarded.
Community Features
The platform also includes food posts, follows, messaging, notifications, and restaurant discovery to create a broader ecosystem around responsible food consumption.
Our core idea is:
Surplus Food → Lower Price → Less Waste
How we built it
We built EcoSort AI as a full-stack web application using Python and FastAPI for the backend and HTML, CSS, and vanilla JavaScript for the frontend.
For AI functionality, we integrated multiple models and services:
- OpenAI GPT-4o for vision-based analysis
- Google Gemini as an additional vision option
- HuggingFace models for AI functionality
flax-community/t5-recipe-generationfor recipe generationsentence-transformers/all-MiniLM-L6-v2for embeddings
The backend exposes API endpoints for waste classification, pantry scanning, and recipe matching.
For data persistence, EcoSort AI supports SQLite and optional MongoDB.
The application is organized into separate modules for vision processing, recipe matching, meal planning, authentication, and data storage, allowing us to extend the platform without rebuilding the entire system.
Challenges we ran into
One of our biggest challenges was making AI classification useful beyond simply producing a label.
A classification such as "plastic" isn't always enough. Different materials and local disposal systems can require different handling, so we needed the system to provide actionable disposal guidance rather than only returning a category.
We also had to design the platform around multiple AI capabilities without making the user experience complicated. Waste classification, pantry scanning, recipe generation, and meal planning all have different workflows, so we focused on keeping the interface straightforward.
Another challenge was building reliable recipe recommendations from incomplete ingredient lists. We addressed this with a combination of AI generation, an existing recipe catalog, embeddings, and fallback search.
Finally, we had to balance the scope of a large platform with the limited development time of a hackathon. We prioritized the features that most directly demonstrate our sustainability goal.
Accomplishments that we're proud of
We're proud that EcoSort AI goes beyond being a simple waste classifier.
We built a platform that connects waste identification, food preservation, recipe generation, meal planning, and surplus sharing into one sustainability-focused system.
We're particularly proud of the AI vision workflow because a user can start with something as simple as a photograph and receive an understandable recommendation about what to do with the item.
We're also proud of integrating multiple AI technologies into a single application while keeping the architecture modular.
Most importantly, EcoSort AI turns the sustainability message into practical actions:
Identify → Understand → Reuse → Share → Reduce Waste
What we learned
We learned that building an AI application is not just about choosing a powerful model. The most important part is turning the model's output into something useful for a real person.
We learned how to integrate different AI services, build APIs around AI workflows, handle model fallbacks, and connect machine-learning functionality with a complete web application.
We also learned that sustainability projects need measurable and understandable outcomes. Instead of simply saying that our project "helps the environment," we designed EcoSort AI around concrete behaviors such as reusing ingredients, reducing unnecessary purchases, correctly sorting waste, and sharing surplus food.
Most importantly, we learned how to turn a broad sustainability problem into a working MVP that people can actually interact with.
What's next for EcoSort AI
Our next goal is to make EcoSort AI more intelligent, localized, and measurable.
We want to add:
- Local disposal rules so recommendations can account for different communities and recycling systems
- Better food-expiry prediction using product and storage information
- Food waste impact tracking to estimate food and resources saved
- Smarter surplus matching between people, restaurants, stores, and nearby organizations
- Computer-vision improvements for more accurate waste and pantry recognition
- Personalized sustainability dashboards showing users how their actions change over time
- Mobile applications for faster scanning and on-the-go use
Ultimately, we want EcoSort AI to become a practical sustainability companion that helps people answer one question every time they are about to throw something away:
"Can this still be used?"
Log in or sign up for Devpost to join the conversation.