About

I’m a research scientist, interested in improving reasoning capabilities of large language models. I’m currently a Member of Technical Staff at Anthropic.

In the past, I’ve worked extensively on multilingual machine translation, unsupervised translation, and multilingual style transfer all while working within the Research team within Google Translate and then Google Brain. This resulted in many interesting improvements to Google Translate, mostly in the form of new language pairs.

Over recent years, however, I’ve been more interested in large language models and spent most of my time at Google (then eventually DeepMind) working on most of the models developed there, including PaLM-1, PaLM-2, Gemini 1.0, Gemini 1.5, etc. Most recently, I worked on improving the reasoning capabilities of language models under the Blueshift team. As an artifact of this work, we developed the math-specialized Gemini 1.5 Pro model.