Google Scholar | Twitter

I am a research scientist at Google DeepMind, where I develop training algorithms for Gemini. My current focus is recursive self-improvement: how can we use Gemini to improve itself? Before that, I spearheaded Google’s effort to detect LLM-generated content, developing a novel watermarking scheme along the way. Since 2018, my research has spanned many aspects of language models, including optimization techniques, efficient architectures and inference, and applications to information retrieval and regression.

Before joining Google, I built recommender systems at Twitter (2016–2018), where my work drove substantial growth in the platform’s active user base. From 2014 to 2016, I built an internet-scale computer vision classification platform at a startup that was later acquired.

I graduated from UC Berkeley in 2013 with degrees in Mathematics and Electrical Engineering and Computer Science (EECS).

A proud Californian, I live in the San Francisco Bay Area. Among other things, I enjoy trail running, mountaineering, classical music, gardening, painting, and fun drives. I am curious about many things, deeply humbled and inspired by nature, and I will rarely turn down an adventure or an opportunity to learn something new. Do not hesitate to reach out.