Assistant Professor, Computing & Mathematical Sciences, Caltech
- Date & time
- Thursday, October 15, 2026 · 11:00 AM
- Location
- Donald Bren Hall 4011
Abstract
Self-Improving machine perception explores how we can build visual systems that do not just perceive the world, but continually improve at doing so. In this talk, I will argue that the next leap in AI will come not only from larger models, but from closed-loop learning systems that generate, critique, and re-curate their own training data. I will present recent work on 3D and video generation, and discuss how self-improving learning can enable more capable world models, more reliable perception, and AI systems that learn in domains where labeled data is scarce.
About the speaker
Dr Gkioxari is an assistant professor in the Computing & Mathematical Sciences department at Caltech. Previously, she was a research scientist at Meta's FAIR team. She completed her PhD at UC Berkeley with Jitendra Malik and her undergraduate studies at NTUA, Greece, where she worked with Petros Maragos. Dr Gkioxari is a Packard Fellow (2025), the recipient of the PAMI Young Researcher Award (2021), a Google Faculty Award (2024), the Okawa Research Award (2024) and the Amazon Research Award (2024). Along with her teammates, she received the PAMI Mark Everingham Award (2021) for the Detectron Library Suite. She was named one of 30 influential women advancing AI in 2019 by ReWork and was nominated for the Women in AI Awards in 2020 by VentureBeat.