Research Engineer, Frontier Safety Mitigations, DeepMind

Research Engineer, Frontier Safety Mitigations, DeepMind Overview

In the role of Research Engineer for Frontier Safety Mitigations at DeepMind, you will play a crucial part in ensuring the safe deployment of AI technologies. Your primary focus will be on mitigating risks associated with model launches, particularly in misuse domains such as cybersecurity and other hazardous areas. You will be responsible for building evaluations, conducting red-teaming exercises, and implementing both in-model and out-of-model mitigations to monitor and address emerging risks.

DeepMind is dedicated to advancing AI for the benefit of society, and the Frontier Safety Mitigation team operates in a collaborative environment that emphasizes support and teamwork. As AI technology evolves, the team is committed to proactively researching and implementing comprehensive safety measures to address potential dangers associated with advanced AI capabilities.

Responsibilities

• Develop advanced classifiers and data pipelines to identify misuse, managing the entire process from automated evaluation to rapid model iteration.
• Create cross-context monitoring systems to detect coordinated harms, innovating signal aggregation methods across different user sessions to pinpoint large-scale attack vectors.
• Implement data-driven, semi-automated account-level response systems to identify, track, and respond to persistent malicious actors using comprehensive signals from production traffic.
• Evaluate and secure agentic AI systems by formulating threat models, establishing testing environments, and deploying effective mitigations against sophisticated hacking attempts and long-term attacks.
• Advance research in automated red-teaming and adversarial robustness, utilizing multi-turn and agentic attacks to systematically identify and address misuse vulnerabilities.

Requirements

• A Bachelor’s degree or equivalent practical experience is required.
• At least 5 years of experience in software development using one or more programming languages.
• Minimum of 3 years of experience in testing, maintaining, or launching software products, along with 1 year of experience in software design and architecture.
• Preferred: A PhD in Computer Science, Machine Learning, or equivalent practical experience, or relevant publications in recognized venues such as NeurIPS, ICLR, ICML, or EMNLP.
• Experience in cybersecurity detection and response, including the development of classifiers and anomaly detection systems at scale.
• Familiarity with transitioning safety defenses or mitigations from research concepts to scalable production systems.
• Knowledge in adversarial machine learning, automated red-teaming, or model interpretability and probes.
• Experience in collaborating on or leading applied machine learning projects, including training, inference, and fine-tuning of large language models.
• Proficiency in using AI coding agents with strong architectural judgment, as well as experience with TPUs and JAX.
• Understanding of AI control, chain-of-thought monitoring, monitorability, and related frontier safety research.

Benefits

• Opportunity to work in a collaborative environment focused on safety and ethical AI development.
• Access to diverse learning opportunities and varied career pathways.
• Engagement in a culture that values teamwork and support.

Degree Requirement: Bachelor’s

Visa Sponsorship May be