Mercor is recruiting fuel-cycle practitioners to write single-turn prompts that sit on the dual-use line — benign, dual-use, and adversarial — and judge whether frontier models handled each correctly against a written policy standard. Every judgment ships with a reference answer and a technical rationale a non-specialist can follow; observed pay is $65–75 per task.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work actually is

You write prompts in your own corner of the fuel cycle — enrichment cascade behaviour, separations chemistry, criticality-safety limits, material balance accounting — and you label each one benign, dual-use, or adversarial before it goes to the model. Then you read what comes back and decide whether the response met a defined policy standard: did it answer a legitimate engineering question completely, or did it hedge into uselessness? Did it refuse something it should have refused, and for the right reason, or did it refuse a routine question because a word in it sounded alarming? Both failures count. The third piece is the reference answer: what a correct response looks like, plus the technical reasoning for why that line falls where it does. That last part is where most of the time goes.

What the screen is looking for

The platform is not testing whether you can recite the fuel cycle. It is testing whether you can tell a routine professional question from one that is fishing — and whether you can say out loud, in writing, what the difference is. Expect follow-ups that push on a specific parameter you named: why that one, at what value does it stop being routine, what does an operator legitimately need it for. Expect at least one scenario where the model's answer is technically correct and still wrong, or wrong in substance but defensibly refused. Writing samples matter here more than in most evaluation work; prior technical writing, published research, or expert witness experience is an explicit signal, and you should have something to link.

Logistics and the boundary you must hold

Remote and asynchronous; work is priced per task rather than per hour, with observed rates of $65–75 per task.
Volume varies with campaign demand — treat it as supplementary rather than a schedule you can count on.
Sustained reading and writing about misuse scenarios in your own field is part of the job. Mercor briefs experts on this in advance and states you can pause or step away without penalty.
The work must not involve or draw on classified or export-controlled material, anything under NDA, or anything subject to prepublication review. Holding such obligations does not disqualify you, but you are expected to disclose them so the work can be scoped around them.

Nuclear Engineer, Fuel Cycle

What the work actually is

What the screen is looking for

Logistics and the boundary you must hold

Interview prep — 10 questions