Nuclear & Radiological Safety Expert for AI Model Audit

In short

You would help audit how AI models answer sensitive questions about nuclear science and radiation. You read a real-world request, judge three answers, and explain what a good answer should look like. The work is remote and paid at 60 to 90 USD per hour on a freelance contract, which may suit experts who know nuclear topics well and enjoy careful, practical review.

What you will do

• Sort each user request into one of three groups: benign, dual-use or adversarial.
• Mark each of the three model answers as safe or unsafe, and explain your reasons.
• Rank the three answers from best to worst.
• For dual-use requests, note what makes the topic sensitive, what legitimate purpose it serves, and who should be able to get an answer.
• Write the answer the model should have given, including cases where a model refused a normal professional question.

What you need to apply

• A relevant degree, such as in nuclear engineering, reactor physics, health physics, radiochemistry or nonproliferation.
• At least 5 years of professional experience in a nuclear or radiological field.
• Experience reviewing technical work against a standard, such as licensing, inspection or peer review, is a strong plus.
• Experience with AI red teaming is also a strong plus.

Who can apply

The role is open to residents of the US, GB, CA, AU and NZ. The posting also names the European Union, so citizens or residents of the EU may apply as well.