Open-Source Intelligent Tutoring System for Programming Exercises in Engineering Education

Implementing control and machine learning algorithms in MATLAB and Simulink is a critical competency in advanced control engineering. While immediate feedback is essential for fostering intuitive understanding, it is traditionally constrained to scheduled exercise sessions or consultation hours. To bridge this gap, this project introduces an open-source intelligent tutoring platform that provides continuous, on-demand feedback. To accommodate diverse solution strategies, the platform employs a hybrid evaluation strategy combining result-based and code-based metrics. This ensures that valid alternative solutions that differ from predefined sample solutions are not misclassified. In case of incorrect solutions, a Large Language Model, contextualized with sample solutions and task classification results, offers auxiliary support for students struggling to initiate or complete tasks. Instructional scaffolding is adaptively adjusted to guide students toward independent problem-solving. We position this platform as a supplementary tool designed to enhance, rather than replace, valuable interactions between students and human tutors. Built on open-source tools, the system is architected for reusability, enabling lecturers across engineering subjects to adapt the framework to their teaching needs easily.