Seizing the means of prediction (robustly)!
I am a postdoctoral researcher at the Bethge Lab at the Tübingen AI Center. My goal is automated scientific discovery, with a focus on:
- Scaling supervision with RL to reduce reliance on manual labeling
- Data‑centric Research: Curating data for pretraining & post‑training (also scalable RL environments)
- Science of benchmarking: Building scalable benchmarks and analyzing current evaluations to better measure progress.
If you're a PhD student interested in collaborating or just want to chat, please fill this form!
Values and research philosophy
- Robust, empirical research: Results which just work in other contexts; avoiding toy setups and unscalable algorithms.
- Simple, principled algorithms: Papers which provide clarity over trial-and-error, validated in real settings — not under the streetlight of convenience.
- Design better incentives via mechanism design, so technology trends toward a fairer world without defaulting to gatekeepers.
Outside of research, I am a techno DJ and am currently learning to shuffle. I advocate for animal welfare, open research culture (JMLR statement, Cost of Knowledge). I believe the underlying open source philosophy, vital for maintaining non-extractive digital institutions (inspiration).