Urban waste pollution often results from inadequate collection and sorting practices. We propose ECO–RACK, a smart waste management system that leverages IoT, embedded systems, and artificial intelligence. Smart bins equipped with ESP32 microcontrollers, ultrasonic sensors, and cameras classify waste in real time using convolutional neural networks (CNNs) and sort it via a servo mechanism. A webcam can capture images of individuals littering, and face recognition is employed to identify repeat offenders under privacy-by-design safeguards. Collected data are transmitted to a central server, where optimized collection routes are computed using a modified Dijkstra algorithm with the OpenStreetMap (OSM) API, prioritizing full bins to reduce trips, fuel, and emissions. The system is designed to handle diverse waste types under varied conditions and to accurately separate recyclable, compostable, and non-recyclable categories. Overall, ECO–TRACK improves sorting, monitoring, and collection efficiency in a scalable, intelligent, and environmentally sustainable manner.