Research Scientist, World Models, DeepMind

Overview

In this role, you will drive research and implementation of new architectures for frontier generative models with real-time performance. You will design and evaluate datasets and collaborate across capability workstreams to scale optimal training recipes. You will contribute across the training stack to land advanced generative and world model capabilities, staying current with diffusion models and generative video research. You join a core modeling team to deliver interactive multimodal models while partnering with researchers and engineers to maintain tight feedback loops and roadmaps.

Responsibilities

• Drive research and implementation of new architectures for frontier generative models with real-time performance
• Design, evaluate, and generate datasets to explore model capabilities
• Collaborate across capability workstreams to scale training recipes
• Contribute across the training stack to land new generative and world model capabilities
• Stay current with diffusion models and generative video architectures
• Engage in significant research, road-mapping and engineering work to deliver world models
• Collaborate closely with researchers and engineers to maintain feedback loops across modeling and data efforts

Key requirements

• PhD in Computer Science, Machine Learning, Statistics, Mathematics, or related technical field, or equivalent practical experience
• 2 years of experience publishing research in machine learning conferences or releasing open/closed-source software models
• 2 years of experience in diffusion models, multimodal generative models, or generative video models

• cross-functional collaboration
• ability to work in interdisciplinary teams
• effective communication of complex research ideas
• diffusion models
• multimodal generative models
• generative video models

…