LectureChat: a question answering assistant for university lectures
Research assistant. Design and development of the retrieval system and its supporting services.
LectureChat answers student questions from the course lectures themselves. What sets it apart from a general assistant is that every answer carries a citation back to the point in the lecture it came from, so the student can verify it.
I also worked on automatic topic segmentation of video lectures, on building a bilingual corpus of university lectures for evaluating the methods, and on the services underneath all of it: search, indexing, segment aware video playback, and monitoring tools.
The system was presented at ECIR 2026, and the segmentation method was published at the ECAI 2023 Workshops.
Why it belongs here: when we talk about adding AI to one of your processes, I do not mean plugging in an off the shelf chatbot. I mean I have built these systems from scratch and I know where they fail.
- Python
- RAG
- Speech recognition
- Search
- Streaming