UIUC AI Teaching Assistant: A Multi-Model System Better Than Google

UIUC AI Teaching Assistant

I built an AI Teaching Assistant for Electrical Engineering at UIUC that runs 11 models in parallel to deliver answers in under two seconds. By combining data from textbooks, lecture videos, and student forums, we created a system that outperforms standard search. I also developed a novel semantic search approach for RLHF using a dataset generated by students, which is now open source on HuggingFace.

"GPT-3 nearly always thinks that GPT-3 is great, which is probably not true and a limitation of this evaluation method."