Research Engineer, World Models, DeepMind

Overview

As a Research Engineer at Google DeepMind, you will architect and implement cutting-edge model architectures and evaluation pipelines to advance frontier world models. You will scale real-time video generation via distillation and optimization, and build scalable data ingestion and filtering to improve training quality. Collaborating with research, data, and product teams, you’ll drive rapid, evidence-based iterations and contribute to open research and production deployments. You will shape scalable, high-impact AI systems that bridge research breakthroughs and real-world applications.

Responsibilities

• Architect and implement next-generation model architectures in collaboration with research teams
• Design and iterate real-world evaluation pipelines to measure and accelerate model capabilities
• Engineer high-performance distillation and optimization for real-time serving at global scale
• Build scalable data ingestion pipelines with automated filtering and rigorous dataset analysis
• Collaborate with modeling, data, and eval teams to scale capabilities and interface with product teams
• Contribute to robust, scalable codebases and AI-driven automation workflows

Key requirements

• Bachelor’s degree or equivalent practical experience
• 5 years of experience developing, training, and deploying diffusion models and generative video architectures
• Experience writing custom GPU/TPU kernels (e.g., JAX or PyTorch) for model serving

• Cross-functional collaboration
• Technical leadership
• Strong communication
• Diffusion models
• Generative video architectures
• Custom GPU/TPU kernels (JAX or PyTorch)

…