AI Safety Experts — English & Assamese

• Red team conversational AI models and agents, including jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation.
• Generate high-quality human data by annotating failures, classifying vulnerabilities, and flagging systemic risks.
• Apply structure by following taxonomies, benchmarks, and playbooks to ensure consistent testing.
• Document reproducibly by producing reports, datasets, and attack cases that customers can act on.

• Bring strong judgment about language and content, with the ability to explain why an AI response is accurate, complete, and appropriate.
• Be rigorous and notice subtle errors, inconsistencies, and gaps.
• Be structured and work to guidelines and quality standards consistently.
• Be communicative, explaining reasoning clearly to technical and non-technical audiences.
• Be adaptable and thrive moving across projects, task types, and customers.

• Experience with adversarial ML, including jailbreak datasets, prompt injection, RLHF/DPO attacks, and model extraction.
• Background in cybersecurity, including penetration testing, exploit development, and reverse engineering.
• Familiarity with socio-technical risk, including harassment/disinformation probing, abuse analysis, and conversational AI testing.
• Skills in creative probing, psychology, acting, or writing for unconventional adversarial thinking.

Project and contracting process: Mercor. Applications continue on the provider's website.