Responsibilities
- Collaborate directly with client security teams in technical workshops to evaluate and strengthen their end-to-end AI-powered attack workflows, from reconnaissance through to sandboxed execution.
- Develop and enhance the exploitation agent’s architecture, focusing on how large language models plan attack sequences, choose and verify exploits, and manage parallel sandboxed operations securely and consistently.
- Improve the cost efficiency of the exploitation pipeline by testing and comparing open-source and sovereign models (e.g., Kimi, GPT-OSS, MiniMax, DeepSeek) against advanced models across reconnaissance, exploitation, and analysis phases; assess accuracy, speed, and cost to guide infrastructure decisions.
- Define and refine the anomaly detection system for runtime environments by identifying high-signal precursor patterns to intrusions or privilege escalation, reducing noise from excessive logs, and designing response automation and criticality-based alert routing.
- Deliver a rapid proof-of-concept to demonstrate early value, such as automated scanning of dependencies or pull requests for vulnerabilities, or a performance comparison between local and frontier models in exploitation tasks.
- Translate technical outcomes into a clear, evidence-based proposal and strategic roadmap, tailored for presentation to senior technical leaders like CISOs or CTOs.
- Ensure all development and sensitive operations remain within secure boundaries—no transfer of intellectual property, configurations, or reconnaissance-enabling data to external AI providers—and comply with regulated gaming certification requirements by excluding uncertified AI from critical runtime systems.
