I am a researcher working on AI safety and alignment. Specifically, I am interested in language models, reinforcement learning and interpretability.
Before, I was a post-doc at University of Bordeaux, working with Nathanaël Fijalkow on Safe AI through formal methods project.
In 2023-2024, I was an ATER (aka teaching assistant) at Université Paris Cité, doing my research at IRIF.
I got my PhD from École Polytechnique, where I worked with Eric Goubault and Jean Goubault-Larrecq on topological models of multi-agent logic and distributed computing.
The best way to contact me is by sending an e-mail to: roman (AT) knzv.me
Here is my CV.
This is how I look like.
