Research Scientist, Multisensor Robotics, DeepMind

Overview

As a Research Scientist at Google DeepMind, you will advance embodied AI by architecting and training models that enable real-world robotic interaction. You will work with cross-disciplinary teams to design, test, and deploy sim-to-real solutions, multimodal models, and efficient on-device architectures. You will prototype concepts, address hardware requirements, and communicate results to diverse audiences. This role offers the opportunity to push the frontier of robotics and AI at scale while contributing to safe, high-impact innovations.

Responsibilities

• Architect and train models to improve robotic capabilities and prototype new concepts
• Tackle sim-to-real transfer, imitation learning, and Vision-Language-Action models
• Optimize models for production using distillation, pruning, and quantization
• Design experiments, deploy promising ideas at scale, and communicate results to stakeholders
• Address hardware requirements for manipulation tasks and real-world deployment
• Collaborate with cross-functional teams to translate research into products or processes
• Share findings with the research community and collaborate with partner institutions
• Set up large-scale tests and iterate on architectures from experiments to production-ready systems

Key requirements

• PhD in Robotics, Computer Science, or Machine Learning, or equivalent practical experience
• 5 years of professional research experience in industrial R&D
• 3 years of experience designing, training, and scaling RL, IL, or multimodal generative architectures (including VLA or video foundation models)
• 2 years of experience optimizing foundation models for on-device latency and memory constraints

• strong communication and presentation abilities
• problem-solving mindset with initiative
• collaboration across cross-functional teams
• reinforcement learning (RL)
• imitation learning (IL)
• multimodal generative architectures (VLA, video foundation models)

…