AWS Builder Center

The Bin Has a Brain. Who Gives It Permission? Building ReLoop with AI and Cedar

What happens when AI-powered waste classification meets open-source authorization? Discover how I built ReLoop using React, FastAPI, PyTorch, Hugging Face Transformers, and Cedar—an open-source authorization engine developed by AWS. This article explores the engineering challenges, real authorization tests, lessons learned, and how responsible AI applications need both intelligent predictions and clearly defined permissions.

student

The Bin Has a Brain. Who Gives It Permission?

Building ReLoop: Where waste classification meets authorization engineering

Imagine holding a dead battery. You know it shouldn't necessarily end up in an ordinary bin, but you aren't sure where it belongs. Now imagine pointing your phone at it and getting a predicted category, a confidence score, and guidance on what to do next.
That is the idea behind ReLoop.
But while building this project, I started thinking about a different question.
If we teach applications to make decisions, how do we control which decisions they're allowed to make?
That question led me beyond machine learning and into authorization engineering with Cedar, an open-source authorization engine originally developed by AWS.
This is the story of how I brought those two ideas together in a student project.

1. The Problem Isn't Just Recognizing Waste

Waste management has an information problem.
People encounter objects made from different materials every day, yet identifying the right disposal route isn't always straightforward. Batteries, cardboard, glass, textiles, and ordinary household waste can require different handling or collection options.
A classifier can help recognize an object, but recognition alone doesn't solve the entire probl
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article