Gabriel Barth-Maron

Gabriel Barth-Maron is a Staff Research Engineer at Google DeepMind, based in London. His work primarily focuses on reinforcement learning, data-efficient learning, multimodal modeling, and training large-scale models. He has made significant contributions to the field of Distributed Reinforcement Learning, particularly through his work on Distributed Prioritized Experience Replay and Distributed Distributional Deterministic Policy Gradients (D4PG), which have demonstrated the effectiveness of using distributed approaches in reinforcement learning.1

Barth-Maron has also been involved in developing tools like Acme, Reverb, and Launchpad to facilitate distributed reinforcement learning research. Recently, he has been working on extending transformer models to multiple modalities, contributing to projects such as Gato—a multi-modal, multi-task, multi-embodiment generalist policy—and is part of the Gemini team at Google DeepMind, working on next-generation large-scale multimodal transformer models.1

He holds a Bachelor of Arts in Mathematical Economics and a Master of Science in Computer Science from Brown University.2 His work has been recognized with awards such as the Outstanding Certification (Best Paper Award) for Gato by Transactions on Machine Learning Research (TMLR).1