The posting
Responsibilities
- Study AI agent tools and coding assistants to find prompt injection and tool-abuse weaknesses.
- Run comparison tests that try to break open-source and proprietary AI models.
- Publish original security research that helps shape enterprise AI safety standards.
- Turn research findings into real-time detection and blocking rules in the product.
Requirements
- 2–10 years as a security or threat researcher focused on AI models rather than only traditional software.
- Proven record of breaking AI agents, tool-calling pipelines, and model boundaries.
- Able to write production detection logic and brief security leaders on new AI threats.
- US citizen or Green Card holder, in San Francisco 5 days a week.
Skills Required
- AI & LLM security
- Red teaming
- Prompt injection defense
- Agent and tool abuse analysis
- Detection engineering
- Security research & writing
Location: San Francisco, on-site 5 days a week. US citizen or Green Card holder only. No sponsorship.



