Remote Adversarial AI Safety Engineer

Mercor; San Francisco (CA); On-site; USD 100,000 - 140,000; Full time

Job summary

Obsidian is looking for professionals in AI red teaming to help identify and document vulnerabilities in AI systems. The role involves working with adversarial inputs and conducting thorough reviews on sensitive AI topics such as bias and misinformation.

Key responsibilities include generating quality human data, structuring tests according to best practices, and delivering clear reports. The ideal candidate has experience in AI adversarial work and excellent communication skills for both technical and non-technical audiences.

Qualifications

• Prior experience in AI adversarial work or cybersecurity.
• Ability to explain risks clearly to both technical and non-technical stakeholders.
• Experience with adversarial ML techniques or cybersecurity methods is a plus.

Responsibilities

• Red team conversational AI models and identify vulnerabilities.
• Generate high-quality annotations and reports.
• Document findings reproducibly and consistently across projects.
• Collaborate with stakeholders to enhance AI safety.

Skills

Red teaming experience

Curiosity and adversarial mindset

Structured approach

Strong communication skills

Adaptability

Job description

Obsidian is looking for professionals in AI red teaming to help identify and document vulnerabilities in AI systems. The role involves working with adversarial inputs and conducting thorough reviews on sensitive AI topics such as bias and misinformation.

Key responsibilities include generating quality human data, structuring tests according to best practices, and delivering clear reports. The ideal candidate has experience in AI adversarial work and excellent communication skills for both technical and non-technical audiences.