Research Scientist, Reinforcement Learning, Deepmind

Google Google · Big Tech · London, United Kingdom

Research Scientist at Google DeepMind focusing on Reinforcement Learning. The role involves proposing and pursuing novel research directions, implementing algorithmic ideas, conducting end-to-end experiments, designing evaluations, and building infrastructure for large-scale research. The scientist will contribute to the wider research community through publications and presentations. The team has a strong track record with significant AI models and is focused on advancing AI development for complex global challenges.

What you'd actually do

  1. Propose and pursue novel research directions by formulating and testing hypotheses.
  2. Implement algorithmic ideas and conduct end-to-end experiments, analyzing results and iterating on findings.
  3. Design evaluations and ablations to answer key questions and drive research decisions.
  4. Build and improve infrastructure to enable research at scale.
  5. Communicate research findings through write-ups, presentations, and publications, while fostering a culture of high standards and constructive feedback.

Skills

Required

  • PhD in Machine Learning or equivalent
  • Experience implementing algorithms in research codebases
  • Experience in reinforcement learning research
  • Contributions to peer-reviewed publications
  • Experience designing and executing end-to-end experiments

Nice to have

  • Advanced RL topics (RL for sequence models, post-training, preference-based learning, agentic systems)
  • Modern research stacks (JAX/Flax or PyTorch)
  • Scaling experiments
  • Experimental judgment
  • Scaling methodologies
  • Evaluation techniques
  • Diagnosing complex failure modes
  • High agency and drive
  • Prioritization skills
  • Initiative

What the JD emphasized

  • PhD in Machine Learning, or equivalent practical experience.
  • 2 years of experience implementing algorithms within research codebases.
  • Experience conducting research in reinforcement learning, including contributions to peer-reviewed publications.
  • Experience designing and executing end-to-end experiments, including setup, analysis, and interpretation.
  • Experience with advanced reinforcement learning topics, such as RL for sequence models, post-training, preference-based learning, or agentic systems.

Other signals

  • DeepMind's RL team
  • building DQN, AlphaGo, Rainbow, AlphaZero, MuZero, AlphaStar, AlphaProof and Gemini
  • advancing AI development
  • pushing the boundaries across multiple domains
  • novel research directions
  • end-to-end experiments
  • design evaluations and ablations
  • build and improve infrastructure
  • peer-reviewed publications
  • advanced reinforcement learning topics
  • RL for sequence models
  • post-training
  • preference-based learning
  • agentic systems