This abstract was part of the 2026 UCSF AI and Education Symposium
Faro addresses a common educational problem in graduate training: interdisciplinary seminars are plentiful, but comprehension is uneven because speakers assume a background that many listeners do not share. Important ideas slip by when a prerequisite concept is missing, and the listener cannot pause the room to catch up. Curiosity gets punished by speed, and cross-field motivation is quietly reduced even in an institution that does an exceptional job convening experts.
Faro is a real-time AI seminar copilot that adapts dense material to each attendee’s unique background and level of understanding. It listens and transcribes in real time, identifies likely knowledge gaps using a dynamic learner profile, and surfaces concise, optional explanations in the moment. Built on UCSF Versa for LLM-based reasoning, Faro generates time-aligned transcripts and grounded clarifications from live or recorded audio. After the talk, it delivers a personalized summary, concept map, and targeted review so listeners leave with a coherent mental model rather than scattered highlights.
Contacts
Tara Pande
Ravishankar Yadav