Frontier AI Safety Research Engineer

United States Digital Space LLC; Greater London; Hybrid; GBP 65,000 - 145,000; Full time

Benefits offered by this job

Hybrid work

25 days annual leave

Annual learning stipend

Conference funding

Pension contribution

Job summary

AI Security Institute in London seeks Research Scientists and Research Engineers to join the Control Red Team. You will design and run ML experiments on frontier models, build adversarial attacks, and evaluate security monitors.

The role offers frontier access, significant compute, and collaboration with leading experts in AI safety and policy. We value candidates who can lead research projects, publish results, and work across research and engineering to reduce real-world AI risk.

Qualifications

• Demonstrated ability to design, build and run ML experiments on frontier models and autonomous research projects.
• Strong software engineering and ML experience: clean, reusable code for ML experiments and evaluation frameworks.
• Ability to critique how an experiment supports safety claims and understand AI safety challenges.

Responsibilities

• Design and run ML experiments on frontier models, including RL and optimization heavy work.
• Develop adversarial attacks and evidence to evaluate monitors and controls.
• Write arguments and discuss results with the team; plan and publish findings.
• Run evaluations of frontier labs' monitors, perform threat modelling and report implications.
• Build tooling and pipelines to accelerate question-to-result delivery.

Skills

ML experiments

Adversarial attacks

LLM fine-tuning

Security analysis

Research infrastructure

Programming

LLM tooling

Tools

Inspect

Job description

AI Security Institute in London seeks Research Scientists and Research Engineers to join the Control Red Team. You will design and run ML experiments on frontier models, build adversarial attacks, and evaluate security monitors.

The role offers frontier access, significant compute, and collaboration with leading experts in AI safety and policy. We value candidates who can lead research projects, publish results, and work across research and engineering to reduce real-world AI risk.