Hyperspectral Textile Recycling ML Platform

Built a full-stack ML platform for textile classification from multilayer and hyperspectral images. Users could upload and label new samples, trigger retraining, and continuously improve detection for new material types. A real-time inference pipeline identified materials on a rolling conveyor chain in production.

Backend in Django; frontend in Angular/TypeScript. Trained Decision Tree, Random Forest, XGBoost, SVM and DNN models, deployed in a Dockerized environment on OpenStack.