AI Safety / Red Team Specialist

Про роль

The safety of the AI analyst is, in our reading, not something one can verify by inspection alone. It requires adversarial testing by someone whose discipline is precisely to find the failure modes a developer would not look for. This role is structured as a contractor engagement, typically project-based evaluation sprints or part-time ongoing engagement.

Обов'язки

• Conduct adversarial testing of the AI analyst chat and the Drusus Daily editorial pipeline.
• Author threat models against the documented standards (NIST AI Risk Management Framework, OWASP LLM Top 10).
• Identify and document the failure modes the platform is exposed to, and recommend the structural mitigations.
• Advise on the safety-posture documentation as the platform evolves.

Кого ми шукаємо

• Substantial demonstrated experience in AI safety, red-team work, or adversarial LLM testing.
• Familiarity with the published literature on language-model safety and the standard failure modes (hallucination, prompt injection, regulatory drift).
• Strong written communication; capable of producing the structured reports the engagement requires.
• Capability to operate with the operational discipline that adversarial work in a financial context requires.