CBRN Safety RLHF Expert

You will help improve AI safety by reviewing prompts and AI-generated responses related to chemical, biological, radiological, and nuclear risks. The work focuses on judging model behavior, identifying safety concerns, and giving clear reference feedback.

• Review prompts and AI-generated responses.
• Classify safety risks in model outputs.
• Evaluate how the model handles CBRN-related requests.
• Rank responses based on their quality and safety.
• Provide expert reference feedback for AI training.
• Take part in red-teaming, which tests how models respond to risky prompts.

This is a part-time contractor role for candidates with intermediate experience. The work requires subject knowledge relevant to CBRN safety and careful judgment when evaluating AI responses.

• Time: Less than 20 hours per week.
• Experience level: Intermediate.
• Relevant knowledge: Chemistry, biochemistry, radiology, nuclear topics, and explosives.
• Key skills: Safety-risk classification, response evaluation, ranking, red-teaming, and expert feedback.