
Nuclear Security Engineer | $65–75/hr | Worldwide Remote | Contract (Task-Based)
Mercor is assembling a panel of nuclear materials and safeguards experts to red-team frontier AI models. In this role, you will help determine whether AI systems can reliably distinguish between legitimate technical inquiries and genuinely dangerous requests — a critical challenge in one of the most sensitive dual-use domains in existence.
Nuclear security sits closer to the dual-use line than almost any other domain. The same material-balance calculation used by an inspector can be repurposed to conceal discrepancies; the same protection assessment that hardens a facility also maps its vulnerabilities. A model that refuses a routine material-accounting question is failing just as badly as one that answers something it shouldn't. Drawing that line reliably requires practitioners — not generalists.
Red-teaming experience is a strong plus across any of these backgrounds.
This is writing-intensive work. Every judgment you make requires a written rationale that a non-specialist can follow. Prior technical writing, published research, or expert witness experience is highly valued — please include a writing sample or link in your application.
You will be reading and writing about misuse scenarios for sustained periods. Mercor briefs all experts on this in advance, and you may pause or step away at any time without penalty.
This work must not draw on classified, export-controlled, or NDA-protected information. If you hold such obligations, you may still be a strong candidate — simply note it in your application and the scope will be adjusted accordingly.