I lead the Gemma post-training team at Google DeepMind. Our mission is to bring intelligence to everyone's device, unlocking local, fast, and private use-cases through the most intelligent-per-byte models.

On top of Gemma, I contributed to the post-training recipe of numerous versions of Bard and Gemini.

Earlier in my career, I was a CIFRE PhD student between Google Brain and INRIA Lille, where I developed new algorithms for reinforcement and imitation learning.