The challenges facing municipal solid waste management include inefficient manual sorting systems, high labor expenses as well as low recycling rates. With automated waste sorting machines, it is possible to achieve a high level of recycling efficiency through accurate sorting of waste materials in real time. In this paper discuss a real-time waste sorting system which combines an image classification model built using Convolutional Neural Network (CNN) with a robotic arm to perform automated recycling operations. The suggested system receives images of waste captured by a camera and sorts them into the preset categories of plastic, metal, paper, and organic waste and activates a robotic arm to put the waste into the right bins. The architectures of lightweight CNN are used to guarantee low latency and real-time performance. The proposed system demonstrates an average classification accuracy of 94.5%, while maintaining a mean inference latency of 70 ms and achieving an overall robotic sorting success rate of 95.5%The suggested strategy will lead to sustainable waste management because it will minimize the human factor and enhance the accuracy of sorting.