The Complaint
Two separate complaints turned out to share one culprit. First: people didn't know how to sort their own waste correctly, and instructions rarely existed in a language they could read. Second: climate education itself rarely reached the communities most exposed to climate change.
The ask was a single tool that could solve both — recognise what someone was holding up to a camera, and teach them something about the planet while it did.
The Investigation
I architected EcoLearn AI as a multilingual Progressive Web App with a Python/FastAPI backend, using Groq-hosted LLaMA-3 alongside Hugging Face models to deliver real-time image recognition for waste sorting.
Point a camera at an item, and the model tells you — in your own language — which bin it belongs in, reasoning over the image rather than a static lookup table.
Around that core, I designed an interactive AI tutor interface with gamified climate quizzes and personalized carbon footprint tracking, to keep sustainability awareness active rather than a one-time scan.
The Verdict
A multilingual PWA that reads a photo of your trash and tells you where it belongs, then keeps you learning through gamified quizzes and a running carbon footprint tally.