Bart Bussmann
Bio
Updated 03/22/26Bart Bussmann is an independent mechanistic interpretability researcher focused on AI safety. He holds an MSc in Artificial Intelligence from the University of Amsterdam (2016-2019) and previously worked as a Machine Learning Researcher at IDLab, University of Antwerp and imec (2019-2022). Since 2023, he has pursued independent alignment research full-time, exploring concepts learned by large language models using sparse autoencoders. His notable contributions include BatchTopK Sparse Autoencoders, Meta-SAEs, and Matryoshka SAEs, with papers published at ICLR and ICML. He has participated in the MATS (ML Alignment Theory Scholars) program and received funding from the Long-Term Future Fund to explore a technical AI alignment research direction.
Community Signal
Updated 03/22/26No endorsements yet.
Links
Updated 03/22/26- Personal Website
- https://bbussmann.com/
- Twitter / X
- LessWrong
- Stuckwork
- EA Forum
- -