I’m a PhD at the Tübingen AI Center, advised by Matthias Bethge and Maksym Andriushchenko. I study how to rigorously measure what LLMs can and can’t do, especially as autonomous agents over long horizons, so we can deploy them safely and reliably.

I co-led PostTrainBench, which measures whether LLM agents can autonomously post-train LLMs, and A Sober Look at LM Reasoning, which showed that many reported reasoning gains shrink under standardized evaluation. Most recently, at MATS with Mary Phuong (Google DeepMind), I trained model organisms for sandbagging: models that deliberately underperform on evaluations.

Earlier, I worked on interpretability. At LASR Labs with Joseph Bloom (UK AISI), we discovered “feature absorption”, a systematic failure mode of sparse autoencoders. At Microsoft Research with Navin Goyal, I studied how representations of harmful concepts change during post-training.

A long time ago, I worked in computational neuroscience at the Max Planck Institute for Biological Cybernetics and EPFL, modeling how visual features map onto neural activity. I hold a dual degree in Computer Science and Biological Sciences from BITS Pilani.