About the work Mercor is assembling a panel of radiological safety experts to red-team frontier AI models. The goal is to test whether a model can correctly judge the misuse potential of a technical request — answering legitimate questions fully while refusing genuinely dangerous ones. You will: Write challenging single-turn prompts in your domain, labelled across three levels: benign, dual-use, and adversarial. Evaluate the responses that come back against a defined policy standard, and judge whether each was handled correctly. Write the reference answer — what a correct response looks like, and the technical reasoning for why. This domain lives on the dual-use line more than most. The same isotope underpins cancer therapy and industrial radiography; the same shielding calculation protects a worker and conceals a source. A model that refuses a routine dose question is failing as badly as one that answers a question it shouldn't. Drawing that line reliably is why we need practitioners rather than generalists. Background we're looking for We're looking for people who have held the licence, not just worked under it — the person of record for acquisition, inventory,