Creating an AI assistant for STEM education

This project is part of the CS-552 “Modern Natural Language Processing” course at EPFL, taught by Prof. Antoine Bosselut.

It aims to develop an AI assistant for STEM education, combining four specialized components — a reward model , MCQA model, retrieval-augmented generation (RAG), and quantized inference. Each component was developed by a member of the team, and I led the development of the reward model.

You can find below the project report.