I lead the Gemini Post-Training effort at Google DeepMind in Mountain View, where my team works on reinforcement learning, reasoning, alignment, evaluation, and unified modeling recipes across all the Gemini models.

I joined Google in 2015 after completing my M.S. in Computer Science at Stanford University, where I worked with Christopher Manning in the Stanford NLP Group. Earlier at Google Research, I worked on Google's Neural Machine Translation (GNMT) system, pioneered single-model zero-shot multilingual translation across 100+ languages in Google Translate, co-created the XTREME cross-lingual benchmark suite, and researched direct speech-to-speech translation and joint speech-text pre-training.

Research Focus

2015 — Present
2023 — Present

Frontier Post-Training, Reasoning & Agents

Post-training all Gemini models since Gemini 1.0 to Gemini 4 Argon.

2019 — 2023

Multilingual Pre-Training, Speech & Cross-Lingual Transfer

Large-scale multilingual foundation models (PaLM 2, mT5/nmT5), joint speech-text representations (SLAM, mSLAM, Mu²SLAM), and cross-lingual evaluation benchmarks (XTREME, XTREME-R, XTREME-S, XTREME-UP).

2015 — 2019

Neural & Zero-Shot Machine Translation

Core architecture and deployment of Google's Neural Machine Translation (GNMT) and the first single-model multilingual NMT system enabling zero-shot translation between unseen language pairs.

Publications

Selected

Selected Talks & Essays

Lectures & Writing

Experience & Education

Chronology
  • 2023 — Now
    Google DeepMind
    Vice President, Research · Gemini Post-Training Lead (Mountain View, CA)
  • 2015 — 2023
    Google Research
    Staff Research Software Engineer · Machine Translation & Multilingual NLP
  • 2013 — 2015
    Stanford University
    M.S. in Computer Science (Artificial Intelligence) · Research Assistant, Stanford NLP Group (Prof. Christopher D. Manning)