AI Alignment Red Team Research Engineer

Aisi; Greater London; Hybrid; GBP 65,000 - 145,000; Full time

Benefits offered by this job

Hybrid work environment

25 days annual leave

Pension contribution 28.97%

Job summary

The AI Security Institute is seeking Research Engineers and Scientists to join the Alignment Red Team in London. You will explore misalignment in frontier models, conduct pre-deployment evaluations, and contribute to public research publications.

You will design and build software tooling for alignment evaluations, collaborate with frontier labs and government partners, and mentor external researchers. Hybrid work and strong research support are provided.

Qualifications

• Ability to work autonomously on complex research projects involving substantial engineering.
• Strong software engineering and ML experience with language models and ML frameworks.
• Experience writing clean, documented research code for ML experiments (e.g. PyTorch).
• Proven ability to work in a team and adapt to needs.
• Impact-driven mindset and high-velocity output.

Responsibilities

• Research methods to automatically search for misalignment in frontier models and loss-of-control risks.
• Build and run alignment evaluations not covered by existing benchmarks.
• Carry out pre-deployment evaluations and report results to frontier AI entities and governments.
• Contribute to public research publications and technical reports.
• Design and build software/tools, including open-source, for improved evaluations.

Skills

Autonomous research

Python for ML

PyTorch

Research code

Team collaboration

Impact-driven

High velocity

Tools

Inspect

Job description

The AI Security Institute is seeking Research Engineers and Scientists to join the Alignment Red Team in London. You will explore misalignment in frontier models, conduct pre-deployment evaluations, and contribute to public research publications.

You will design and build software tooling for alignment evaluations, collaborate with frontier labs and government partners, and mentor external researchers. Hybrid work and strong research support are provided.