Control Red Team – Research Engineer/Research Scientist

Responsibilities

We are representing a well-funded public AI research organisation seeking a Research Engineer or Research Scientist to join their control red-team initiative. This is a unique opportunity to shape a newly formed team focused on evaluating AI safety measures at scale.

Your work will span two core tracks:

Research track: You will design and run machine learning experiments that test how effectively control measures—monitors, sandboxes, permission systems—detect and prevent harmful model behaviour. This includes conceptually difficult questions: How do we measure recall against threats we have not yet observed? What evidence supports safety claims, and what are the limits of that evidence? Day-to-day, you will design ML experiments (including reinforcement learning and optimisation-heavy work), build adversarial attacks, and produce written arguments that stand up to scrutiny. The team intends to publish this research.

Testing track: You will conduct hands-on evaluations of real control systems deployed by frontier organisations. This includes threat modelling realistic AI attacker scenarios, breaking monitors and sandbox infrastructure, performing security analyses, and producing decision-relevant reports for both the organisations and government stakeholders.

Across both tracks, you will build tooling and experimental pipelines that enable rapid iteration at a high quality bar. You will make heavy use of large language models to automate attack, evaluation, and analysis loops—leveraging model improvements to accelerate your own work. You will also build and operate the infrastructure required to train and serve models at the scale your experiments demand.

What We're Looking For

Essential requirements:

• Demonstrated ability to design, build, and run ML experiments on frontier models, working autonomously on complex research projects with substantial engineering. This includes black-box work (API-based evaluations and attacks) and ideally some white-box work such as fine-tuning open-weight models.
• Strong software engineering and ML experience: writing clean, documented, reusable code for machine learning experiments beyond one-off research scripts, including experience with LLM fine-tuning and inference frameworks or evaluation frameworks.
• The ability to understand and critique how an experiment does and does not support a safety claim, including an understanding of why AI safety and control are hard problems, or a clear appetite to get up to speed quickly.
• Impact-driven mindset and collaborative team player: motivated by the work that most reduces risk rather than what is superficially impressive, flexible about what needs doing, and high velocity with a high-quality bar for outputs.

Highly desirable (you do not need all of these):

• A working model of how frontier AI companies deploy models internally: ML infrastructure, development practices, the kinds of experiments they run, and where security weak points exist.
• Exceptional red-teaming mindset—instinctively finding the path a capable adversary would actually take, whether against a model, a monitor, or a sandbox.
• Experience with ML optimisation: reinforcement learning, supervised fine-tuning, evolutionary methods, or similar. Experience optimising hard against a defined metric and making careful measurement choices.
• Strong written communication and argumentation: high-quality research write-ups where the reasoning, not just the result, is the point.
• Willingness and ability to construct and defend arguments for safety claims, and to think about which claims are worth making in the first place.
• Experience building or operating ML research infrastructure at a large organisation: GPU management, running experiments at scale, securing evaluation environments.
• Experience in cybersecurity or security analysis, including attacking LLM-based applications and agent scaffolds.
• Familiarity with the AI control and adversarial ML literature, and existing relationships with researchers working on control at labs or in the wider safety community.
• Participation in an AI safety research or fellowship programme, or equivalent evidence of independent research output.
• Broad evidence of strong mathematical, scientific, or analytic ability (for example, highly competitive courses or programmes, or olympiad-level results).
• Proficient use of LLM coding tools and agents.

We are less interested in credentials as such: a first-author conference paper or a CS degree is welcome evidence, but neither is required, and neither substitutes for the signals above.

How We Work With You

Get A Job.ai represents candidates applying to this confidential AI safety research organisation. Here is how the process works:

• You apply through the Get A Job.ai platform.
• Our talent team screens your application and conducts an initial conversation.
• We submit qualified candidates to our client for consideration.
• Do not contact the client directly; all communication flows through Get A Job.ai.

The interview process may vary by candidate, but typically includes technical proficiency tests, discussions with a cross-section of the team, conversations with your team lead, and a final conversation with senior leadership.

Candidates should expect some or all of the following stages: initial assessment, initial screening call, technical assessment, behavioural interview, research interview, and final interview with senior leadership.

Pay

This role is open across several seniority levels. Our client offers competitive compensation benchmarked to role scope and relevant experience. Most offers fall between £65,000 and £145,000, made up of a base salary plus a technical allowance. An additional 28.97% employer pension contribution is paid on the base salary.

The full range of salaries:

• Level 3: £65,000–£75,000
• Level 4: £85,000–£95,000
• Level 5: £105,000–£115,000
• Level 6: £125,000–£135,000
• Level 7: £145,000

Additional benefits include: hybrid working with a modern central London office, at least 25 days' annual leave plus 8 public holidays, generous paid parental leave, 5 days off and stipends for learning and development, funding for conferences and external collaborations, opportunities to publish and collaborate externally, and pre-release access to multiple frontier models with ample compute.

Successful candidates must undergo a criminal record check and obtain baseline personnel security standard (BPSS) clearance. There is a strong preference for eligibility for counter-terrorist check (CTC) clearance.

Equal Employment Opportunity: Get A Job.ai is committed to equal opportunity employment. We evaluate qualified applicants without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, veteran status, or any other protected characteristic.