I'm a researcher at Cursor working on RL for the Composer models. Previously, I was a PhD student at Princeton University advised by Danqi Chen. Prior to that, I was an undergraduate at the University of Cambridge, where I was fortunate to work with Adrian Weller. I grew up in Wolfenbüttel, Germany.
My PhD research was centered on building better language models, usually by focusing on simple methods and training data. This includes work on selecting pre-training data [1], optimizing data mixtures [2], careful data repetitions [3], and long-context recipes [4].
I was also part of the team that built SWE-bench and SWE-agent.
Google Scholar/ GitHub/ Twitter/ LinkedIn
(* indicates equal contribution)
Building Better Language Models With Data Curation PhD Thesis 2026
Composer 2 Technical Report Technical Report 2026
Olmo 3 Technical Report 2025
SWE-smith: Scaling Data for Software Engineering Agent NeurIPS Datasets & Benchmarks 2025 (Spotlight)
Finding Transformer Circuits with Edge Pruning NeurIPS 2024 (Spotlight)
QuRating: Selecting High-Quality Data for Training Language Models ICML 2024 (Spotlight)
Language Models as Science Tutors ICML 2024
SWE-bench: Can Language Models Resolve Real-World GitHub Issues? ICLR 2024 (Oral)
Learning Transformer Programs NeurIPS 2023 (Oral)