I am a MATS Research Fellow working with Roger Grosse on training data attribution for large language models.
My broader interests include optimization, statistical learning theory, interpretability and efficient LLM inference.
- Google Scholar
- GitHub
- London, UK
News
- Dec 2026. I will present Efficient and Accurate Causal Validation of Training Data Attribution at the NeurIPS 2026 ATTRIB workshop (code, slides).
- Oct 2026. Masked Language Flow Models accepted to the COLM NonAR-LM Workshop 2026.
I taught a hands-on Introduction to LLMs course at Oxford and Imperial College London.
Previously, I was a postdoctoral researcher at Oxford Statistics and UCLouvain. I completed my DPhil in Mathematics at Oxford in 2021, supervised by Jared Tanner, working on low-rank and sparse recovery (thesis).