EcoHabit

Waste on a college campus is generated continuously and sorted inconsistently. Knowing what a given item is — and what should happen to it next — is the step where most circular economy efforts stall.

EcoHabit approaches this as a platform problem: one system spanning students on mobile, services in the backend, and administrators on the web.

Build a multi-app platform for college campuses combining a Flutter mobile app, a NestJS backend, a Python / FastAPI AI service and a Next.js admin dashboard.

Use an image classification model to identify waste items from a photograph, then turn that classification into disposal guidance and upcycling suggestions.

Apply Redis caching within the platform.

• Mobile — a Flutter application as the student-facing entry point.
• Backend — a NestJS service coordinating the platform.
• AI service — a Python / FastAPI service running MobileNetV2 image classification.
• Admin — a Next.js dashboard for administration.
• Caching — Redis.

• MobileNetV2 used to classify images into eight waste categories.
• A classification pipeline running image → MobileNetV2 → classification → confidence threshold → disposal guidance → upcycling suggestions.
• Redis caching used within the platform.

Results and detailed analysis available in the project repository.

• A confidence threshold is what separates a demo classifier from a system that can decline to answer rather than guide someone wrongly.
• Splitting the AI service out from the main backend lets the model evolve independently of the rest of the platform.
• Coordinating four applications around one model means the contract between them matters more than any single codebase.