Remote AI Safety Evaluation Specialist

mercor; San Francisco (CA); Hybride; GBP 62 000 - 73 000; Temps partiel

Résumé du poste

Mercor connects elite creative and technical talent with leading AI research labs and offers a remote AI Safety Practitioner role. You will evaluate AI-generated responses for safety, factual accuracy, policy compliance, and overall quality, supporting RLHF, SFT, and AI safety benchmarking.

You'll review content across sensitive domains, identify unsafe outputs, and provide structured feedback to improve model alignment and safety performance.

Qualifications

• Bachelor’s degree or higher in a related discipline.
• 5+ years of professional experience in AI Safety, Trust & Safety, journalism, public policy, scientific research, security, or a related field.
• Excellent written English, critical thinking, and analytical reasoning skills.
• Ability to consistently evaluate nuanced and policy-sensitive scenarios.

Responsabilités

• Evaluate AI-generated responses for safety, factual accuracy, policy compliance, and overall quality.
• Review content involving misinformation, political persuasion, self-harm, violence, cyber, biosecurity, and other sensitive domains.
• Apply and refine evaluation rubrics for RLHF, SFT, and AI safety benchmarking.
• Identify unsafe outputs, hallucinations, reasoning failures, and policy violations.
• Provide structured feedback to improve model alignment and safety performance.
• Collaborate with AI researchers and safety teams on ongoing evaluation initiatives.

Connaissances

Written English

Critical thinking

Analytical reasoning

Policy evaluation

Attention to detail

Formation

Bachelor's degree or higher in related field

Description du poste

Mercor connects elite creative and technical talent with leading AI research labs and offers a remote AI Safety Practitioner role. You will evaluate AI-generated responses for safety, factual accuracy, policy compliance, and overall quality, supporting RLHF, SFT, and AI safety benchmarking.

You'll review content across sensitive domains, identify unsafe outputs, and provide structured feedback to improve model alignment and safety performance.