AI Red Team Engineer

AI Red Team Engineer Overview

Apollo Research is seeking an AI Red Team Engineer to join our team in either London or San Francisco. This role is pivotal as we develop Watcher, a monitoring tool for coding agents, aimed at translating computational processes into scalable safety measures. The AI Red Team Engineer will be responsible for establishing the practice of red-teaming AI monitors, focusing on both Watcher’s defenses and the monitoring systems of frontier labs.

In this position, you will actively hunt for untested attack surfaces and convert your findings into actionable fixes. You will collaborate closely with the CEO, Marius, who currently leads monitoring efforts, as well as control researchers and product engineers. This role is ideal for individuals who think like attackers and wish to leverage their adversarial insights to enhance AI monitoring systems. You will be part of a small team, allowing you to significantly influence both the team dynamics and technological direction, with opportunities to take on responsibilities quickly.

Key Responsibilities

• Design and run red-teaming campaigns that integrate severity-graded failure-mode injections into real trajectories, static monitoring benchmarks, and dynamic off-policy control red-teaming.
• Identify novel attack surfaces that existing monitors have not yet been tested against.
• Build and maintain automated red-teaming pipelines to conduct large-scale attacks on monitors, moving beyond one-off manual probing.
• Design iterative adversarial red-team/blue-team exercises, collaborating with the blue-team to escalate attack difficulty as monitors improve.
• Track research literature and real-world incidents related to agent failure modes, monitor evasion, and adversarial robustness.
• Translate each campaign into specific, actionable recommendations for monitor developers, focusing on intervention timing, user authorization, trust boundaries, and monitor visibility.
• Document campaign findings in external publications and internal reports.
• Feed findings back into Watcher’s monitor development and Apollo’s blue-teaming research to enhance product resilience.

Job Requirements

• Must have at least 2 years of experience in offensive security, adversarial ML, or red-teaming AI systems.
• Strong experience with AI coding agents, including extensive use, comparison, or direct development of frontier coding agents.
• Experience in designing and executing structured adversarial testing campaigns.
• Proficient in Python programming.
• Strong written communication skills for producing clear and credible publications and campaign reports.
• Demonstrated ability to work independently on open-ended adversarial problems.

Benefits

• Market competitive salary with equity options.
• Salary range: $182,000 – $238,000 in San Francisco; £122,000 – £160,000 in London, with plans for salary increases.
• Flexible work hours and schedule.
• Unlimited vacation and sick leave.
• Up to 6 months of paid parental leave.
• Comprehensive health, dental, and vision insurance.
• Retirement savings with competitive employer matching.
• Lunch, dinner, and snacks provided on workdays.
• Paid work trips, including staff retreats and conferences.
• A yearly professional development budget of $1,000.
• Relocation support and visa fee assistance if applicable.

Working Pattern

This is a full-time, onsite role based in either our London or San Francisco office, with flexible working hours and arrangements available.

About the Team

The product team comprises research scientists and product engineers, with the CEO advising the team. You will also collaborate with other software engineers and researchers, as we aim to utilize our products internally for research purposes.

About Apollo Research

Apollo Research focuses on the risks associated with AI capabilities, particularly concerning loss of control and deceptive alignment. We work on the science of scheming, detection, and mitigation, collaborating with leading AI companies to test their models and contribute to the understanding of AI risks. Our culture promotes truth-seeking, goal orientation, constructive feedback, and a friendly atmosphere.

How to Apply

Please complete the application form with your CV. A cover letter is not required, but you may share links to relevant work samples. The interview process includes a screening interview, a take-home test, three technical interviews, and a final interview with the CEO. We do not conduct leetcode-style coding interviews, and you may use AI tools for the take-home test.

Degree Requirement: Not Specified

Visa Sponsorship Promising