EDSGR: Eindhoven Digital Scientific computing Guidance Repository
Background and general information
As generative AI tools such as ChatGPT and Copilot become more capable of producing functional code, programming education faces a new challenge. While these tools can support learning, they may also lead students to bypass core learning objectives such as understanding abstraction, syntax, and problem-solving strategies. For students to use AI responsibly and effectively, they must first develop foundational programming skills that enable them to assess and refine AI-generated code critically.
To support this, the EDSGR project proposes the development of a dedicated Large Language Model tailored to programming education at TU/e. Unlike general-purpose tools, this model will be specifically aligned with the course content and learning goals, offering support without directly generating code. The tool aims to assist students by prompting reflection and guiding problem-solving rather than providing ready-made answers. At the same time, it offers educators insights into common learning difficulties by analyzing patterns in student questions and interactions. Initially, it will be piloted in two core Mechanical Engineering programming courses and may be extended to other departments depending on results. The initiative reflects a growing need to combine AI support with pedagogical oversight in a way that enhances both learning and teaching.
Goal or aim of the project
The primary aim of this project is to develop, implement, and evaluate EDSGR. A Large Language Model designed to support programming education without undermining core learning objectives. The tool is intended to function as a virtual tutor that helps students navigate complex concepts, receive immediate guidance, and strengthen their ability to think critically about code. By avoiding automatic code generation, EDSGR promotes active learning and encourages students to reflect on their approach rather than rely on automated solutions.
For instructors, EDSGR offers a complementary benefit: it provides real-time insight into student progress and identifies topics where learners struggle most. This enables teachers to tailor their instruction more effectively and manage large cohorts with greater ease. The tool will be piloted in two key Mechanical Engineering courses; 4CA10 and 4EM30, and evaluated through surveys, interviews, and performance comparisons between different student groups. Feedback will inform iterative improvements to the tool and teaching
strategies. If the pilot is successful, the project plans to extend the tool to other programming-related courses and potentially across departments. The integration with the Alexandria platform ensures scalability and long-term sustainability within TU/e’s broader digital learning ecosystem.
For more information, please contact:
