AI Safety Mitigations Engineer: Frontier Models

Lilasciences; Greater London; On-site; GBP 168,000 - 252,000; Full time

Benefits offered by this job

Bonus potential

Equity

Job summary

Lila Sciences is building Scientific Superintelligence to solve humankind's greatest challenges. We are hiring engineers to develop and test safeguards that reduce societal risks from increasingly capable scientific AI, spanning model training, deployment and evaluation.

You will collaborate with domain scientists, risk researchers and engineering teams to turn risk findings into effective interventions, applying safety-focused evaluations across lab-connected workflows.

Qualifications

• Hands-on experience developing safety interventions for frontier models.
• Depth in at least one area: post-training safeguards, model classifiers, or abuse monitoring.
• Experience collaborating with model teams to investigate failures and implement improvements.
• Strong research-engineering skills to turn ambiguous problems into maintainable implementations.

Responsibilities

• Translate risk assessments into concrete mitigation designs and testable requirements.
• Develop safeguards spanning training, classifiers, monitoring, access controls, and runtime interventions.
• Build experiments to measure risk reduction and robustness.
• Work with model-training, post-training, inference and platform teams to integrate mitigations.
• Investigate regressions, distribution shifts, and misuse patterns to guide development.

Skills

Safety interventions

Model safety

Research engineering

Cross-team collaboration

Job description

Lila Sciences is building Scientific Superintelligence to solve humankind's greatest challenges. We are hiring engineers to develop and test safeguards that reduce societal risks from increasingly capable scientific AI, spanning model training, deployment and evaluation.

You will collaborate with domain scientists, risk researchers and engineering teams to turn risk findings into effective interventions, applying safety-focused evaluations across lab-connected workflows.