This paper presents a web-based and machine learning–driven assignment submission and feedback system designed to improve academic assessment processes. The system addresses two major challenges: the manual effort involved in conducting and evaluating assignments, and the lack of detailed, meaningful feedback provided to students beyond simple scores. SmartAssign is developed using the MERN stack (MongoDB, Express.js, React.js, Node.js) integrated with Python for implementing machine learning functionalities. It supports two primary user roles—teachers and students. Teachers can upload student data via Excel sheets, create time-bound MCQ-based assessments across multiple topics, and analyze class performance through an interactive dashboard. Students can register only if their IDs are pre-listed, access active tests, submit responses within deadlines, and receive automated feedback along with their scores. The system employs a K-means clustering algorithm to generate topic-wise feedback and categorize student performance based on historical attempt data. For new users without prior data, a rule-based fallback mechanism ensures feedback generation from current attempts. Experimental testing demonstrates accurate automated scoring, reliable role-based access control, and effective ML-based feedback generation, highlighting SmartAssign’s potential to enhance efficiency, scalability, and learning outcomes in modern education systems.