Inspiration
Most people throw waste away without thinking. We wanted to show what that actually costs in years, in tons, in real environmental damage.
What it does
WasteLens analyzes any waste item from a photo, text, or both. It returns the correct disposal method, what happens if disposed incorrectly, verified statistics, and a chain of environmental harm.
How we built it
Single-page HTML/CSS/JS frontend. Groq API with Llama 4 Scout for multimodal image analysis and Llama 3.3 70B for text-only inputs. Deployed on Vercel with a serverless API proxy to keep the key secure.
Challenges we ran into
Getting consistent structured JSON from the AI across different waste types was tricky. We also spent time making the image input work reliably across devices. Accomplishments that we're proud of The chain of harm section showing how one wrong disposal cascades from local to global impact turned out more powerful than we expected.
What we learned
Multimodal AI can extract meaningful context from a simple photo. The quality of the system prompt determines the quality of the output far more than the model itself. What's next for WasteLens Location-based disposal points, so the app can tell you not just how to dispose of something but exactly where to take it.
Log in or sign up for Devpost to join the conversation.