Research Scientist/Engineer, Autonomous Security, DeepMind

• Design and iterate on post-training methodologies for Gemini, including reinforcement learning from execution feedback, multi-step trajectory reward modeling, and supervised fine-tuning on curated security datasets.
• Build scalable simulated environments and synthetic data pipelines to execute complex security workflows and generate high-signal training trajectories at scale.
• Architect autonomous agent harnesses capable of multi-step planning, tool orchestration, and decision-making under adversarial conditions.
• Develop end-to-end evaluation benchmarks that measure frontier capability limits and trajectory fidelity across realistic offensive, defensive, and vulnerability tasks.
• Transition research innovations into Google's internal defenses and customer products, while contributing to technical reports and ML and security publications.

Minimum qualifications:

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

Preferred qualifications:

• Experience architecting autonomous agent systems, focusing on multi-step planning, tool orchestration, reward modeling, and reinforcement learning.
• Experience building scalable simulated execution environments, evaluation harnesses, or synthetic data pipelines for training and benchmarking foundation models.
• Deep domain expertise in one or more specialized cybersecurity areas, such as vulnerability discovery and automated patching, offensive security operations (red teaming), or advanced threat detection.
• Track record of published research at ML or cybersecurity venues, or a demonstrated history of deploying AI systems in production security environments.

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