Remote AI Safety Red Team Engineer

Obsidian; San Francisco (CA); Remote; USD 100,000 - 150,000; Full time

Job summary

Obsidian is looking for data experts to join a red team that probes AI models with adversarial inputs. This role requires fluency in both English and Punjabi, and involves tasks such as testing conversational AI for vulnerabilities and generating reports.

Ideal candidates will have experience in adversarial AI and cybersecurity, emphasizing the importance of AI safety. Work will be structured, with clear guidelines and wellness resources provided for sensitive topics.

Qualifications

• Fluent in English and Punjabi with native fluency required.
• Experience in AI adversarial work, cybersecurity, or socio-technical probing.

Responsibilities

• Red team conversational AI models through various misuse techniques.
• Generate high-quality human data: annotating failures and classifying vulnerabilities.
• Document reproducibly by producing reports and datasets.

Skills

Red-teaming experience

Curiosity and adversarial mindset

Strong communication skills

Adaptability across projects

Job description

Obsidian is looking for data experts to join a red team that probes AI models with adversarial inputs. This role requires fluency in both English and Punjabi, and involves tasks such as testing conversational AI for vulnerabilities and generating reports.

Ideal candidates will have experience in adversarial AI and cybersecurity, emphasizing the importance of AI safety. Work will be structured, with clear guidelines and wellness resources provided for sensitive topics.