Autonomous Security AI Research Scientist

DeepMind Technologies Limited; Mountain View (CA); On-site; USD 174,000 - 252,000; Full time

Job summary

DeepMind Technologies Limited is seeking a Research Scientist in Mountain View to design post-training methodologies for Gemini, including reinforcement learning from execution feedback, and to build scalable environments for security-focused research. You will architect autonomous agent systems and develop end-to-end evaluation benchmarks across offensive, defensive, and vulnerability tasks.

You will contribute to translating research into defenses and customer products, while publishing

Qualifications

• PhD in CS, cybersecurity, ML, or related field, or equivalent practical experience.
• 4 years of Python and ML framework experience (PyTorch, JAX, TensorFlow) training, fine-tuning, evaluating foundation models.
• 3 years applying ML or automated reasoning to cybersecurity, systems, program analysis, or code generation.
• 2 years in LLM post-training, including supervised fine-tuning, execution-feedback RL, and multi-step trajectory reward modeling.

Responsibilities

• Design and implement post-training methodologies for Gemini, including reinforcement learning from execution feedback and multi-step trajectory reward modeling.
• Build scalable simulated environments and synthetic data pipelines for complex security workflows and high-signal training trajectories.
• Architect autonomous agent harnesses with multi-step planning, tool orchestration, and decision-making under adversarial conditions.
• Develop end-to-end evaluation benchmarks measuring frontier capabilities and trajectory fidelity across offensive, defensive, and vulnerability tasks.
• Transition research into Google's internal defenses and customer products, and contribute to technical reports and publications.

Skills

Python

ML frameworks (PyTorch/JAX/TF)

LLM post-training

Reinforcement learning in security/AI

Education

PhD in Computer Science / related field

Tools

PyTorch

JAX

TensorFlow

Job description

DeepMind Technologies Limited is seeking a Research Scientist in Mountain View to design post-training methodologies for Gemini, including reinforcement learning from execution feedback, and to build scalable environments for security-focused research. You will architect autonomous agent systems and develop end-to-end evaluation benchmarks across offensive, defensive, and vulnerability tasks.

You will contribute to translating research into defenses and customer products, while publishing