You will lead the development and review of software security training tasks for AI coding models. The work turns real-world vulnerabilities, including CVEs, security advisories, and bug reports, into realistic audit and patch tasks across multiple programming languages.
You will combine security research, vulnerability reproduction, benchmark development, automated testing, technical review, and evaluation of whether AI systems can understand, exploit, and securely fix software vulnerabilities.
- Plan batch deliveries, assign work across pod members, and track progress against phase targets.
- Monitor production, identify bottlenecks, and keep tasks moving through each stage.
- Review tasks for quality, security, reproducibility, and evaluation standards.
- Calibrate pod leads and trainers on CWE classification, verifier strength, task difficulty, and acceptance criteria.
- Use throughput and quality data to forecast staffing needs and document risks, findings, and delivery plans.
- Lead status reporting and clarify changes to scope, task volume, and difficulty.