UG05 – OPT Mentor – A serverless web AI Python Tutor

As the demand for e-learning continues to grow (14% GAGR between 2023 and 2032), it becomes increasingly difficult to provide expensive computing resources, especially the GPUs to support LLMs in teaching. It is also challenging to secure a centralized server, especially for programming tools such as OPT that allow students to execute arbitrary code. This complexity often leads to restricted access and functionality, thereby limiting the effectiveness of these tools. The project aims to develop an advanced version of the Online Python Tutor (OPT), named OPT Mentor. It incorporates OPTLite and WebLLM to offer students detailed explanations of error messages in context. It uses students’ own GPU to provide AI-powered suggestions for enhancing their code directly through a web browser without a network. Additionally, OPT Mentor is designed to integrate effortlessly with JupyterLite, allowing students to save and revisit their solutions in Jupyter notebooks anytime.