I’m an AI researcher focused on LLM agents. Previously, I led the Large Language Models Lab at ServiceNow Research.

In 2022, I co-founded BigCode with Hugging Face and co-led it, from overall strategy and technical direction to day-to-day operations. This open-science collaboration grew to more than a thousand researchers and released The Stack, StarCoder, and StarCoder2—the last in collaboration with NVIDIA. Alongside these releases, we set new standards for responsible data governance at scale, including allowing developers to opt their code out of the training data.

Earlier, I co-invented FiLM, a conditioning layer for neural networks, and earned my PhD at Mila under Aaron Courville. I’ve also worked at Facebook AI Research, Twitter, and INRIA.

Selected work: FiLM · The Stack · StarCoder · StarCoder2