Research Scientist/Engineer, Frontier Reasoning, DeepMind

Overview

You bridge research and production to advance frontier reasoning and agentic systems. You will build scalable distributed post-training pipelines and simulation loops, enabling Gemini models to solve complex multi-step problems autonomously. You’ll run rigorous experiments to identify bottlenecks and communicate findings clearly. You’ll maintain high code quality and guide architectural health across RL and modeling codebases. This role offers impact through production-ready research that shapes next-generation AI capabilities.

Pay / Benefits

• bonus target
• equity
• comprehensive benefits

Responsibilities

• Architect and optimize distributed post-training pipelines and agent-environment simulation loops
• Design and run rigorous ablations to diagnose performance bottlenecks
• Maintain code quality and architectural health across shared RL and modeling codebases
• Land research breakthroughs into production features for Gemini releases

Key requirements

• 4 years of experience building, scaling, and debugging ML models with DL frameworks (JAX, PyTorch, or TensorFlow)
• Experience in at least one core area: RL, SFT/RLHF/RLAIF, Agentic Tool-Use, or Inference-Time Search

• collaborative mindset
• clear written communication
• rapid experimentation with rigor
• distributed systems
• post-training pipelines
• asynchronous agent-environment simulation loops

…