Frédéric Berdoz

I am a third-year PhD student at ETH Zurich under the supervision of Prof. Dr. Roger Wattenhofer. I study how to design AI systems that reliably reflect human intentions and values, exploring topics such as theoretical safety guarantees for AI-run governments, preference alignment in LLMs at inference, political opinion alignment for digital twins, and the distributional alignment of AI-generated content across media streams.

Prior to joining ETH, I worked at the fast-growing startup Neural Concept as a technical project manager, where we applied geometric deep learning to solve real-world engineering problems.

I hold a Master’s degree in Computational Science and Engineering from EPFL. For more details, you can find my CV here.