Nuclear AI Safety Red-Teaming Engineer

Obsidian; Greater London; Remote; GBP 70,000 - 110,000; Full time

Job summary

Mercor is assembling a panel of nuclear domain experts in Greater London to red-team frontier AI models. The aim is to test whether a model can correctly judge misuse potential by answering legitimate questions fully while refusing dangerous ones.

You will write prompts across benign, dual-use, and adversarial categories, evaluate model responses, and draft the reference answer with technical reasoning. This is a writing-intensive, policy-focused role requiring clear rationale for every judgment.

Qualifications

• Safeguards and nonproliferation experience signals suitability.
• Strong technical writing and domain knowledge required.
• Ability to produce clear rationales for each judgment.

Responsibilities

• Write challenging prompts across benign, dual-use, and adversarial levels.
• Evaluate model responses against a defined policy standard.
• Draft reference answers and technical reasoning for model outputs.

Skills

Red-teaming experience

Nuclear domain knowledge

Education

Nuclear engineering background

Job description

Mercor is assembling a panel of nuclear domain experts in Greater London to red-team frontier AI models. The aim is to test whether a model can correctly judge misuse potential by answering legitimate questions fully while refusing dangerous ones.

You will write prompts across benign, dual-use, and adversarial categories, evaluate model responses, and draft the reference answer with technical reasoning. This is a writing-intensive, policy-focused role requiring clear rationale for every judgment.