Research Scientist, Human Data, Robotics, DeepMind

Overview

As a Research Scientist at Google DeepMind, you design, implement, and evaluate large robotic models and algorithms, advancing capabilities of robotic agents. You work with a large, fast-paced, cross-functional team to translate research ideas into scalable software and experiments. You develop methodologies to incorporate scalable data sources, including human data, into robotics foundation models and explore learning from simulation, reinforcement learning, and multimodal modeling. You report findings clearly and contribute to both internal teams and the wider research community while pushing dexterity and generalization at scale.

Responsibilities

• Design, implement, train and evaluate large models and algorithms for robotic agents
• Write software to implement research ideas and iterate quickly with a collaborative team
• Develop methodologies and experiments for incorporating scalable data sources, including human data, into robotics foundation models
• Engage in diverse research areas including learning from simulation, reinforcement learning, learning from demonstrations, vision-language-action models, transformers, video generation, robot control, and humanoid robots
• Report and present research findings clearly to internal and external audiences
• Setup large-scale tests and deploy promising ideas quickly and broadly, managing deadlines and deliverables
• Prototype implementations and design new architectures to solve real-world robotic problems
• Contribute to the wider research community by sharing findings and collaborating with universities and institutes
• Address data capture device design, datasets, and modeling approaches to leverage human action data for dexterity and generalization

Key requirements

• PhD in Computer Science, a related field, or equivalent practical experience
• 2 years of experience with reinforcement learning and imitation learning
• Experience with vision, vision-language, video, and other multimodal models
• Experience with multimodal generative modeling, training and inference

• collaboration
• clear communication
• adaptability
• reinforcement learning
• imitation learning
• vision, vision-language, video, and multimodal models

…