Principal Security Researcher – AI (Cybersecurity LLM Post-Training, Evals, and Environments)
Core
Advancing cybersecurity capabilities of large language models and autonomous AI agents through security research, rigorous evaluation, and applied LLM post-training.
Role type
Principal AI Researcher (Cybersecurity LLM Post-Training, Evals, and Environments)
Builds
Production-grade security data, realistic training/evaluation environments, reliable graders, and post-training methods for cybersecurity tasks.
Domain
Cybersecurity + Machine Learning
Deliverable
production ML models
Required skills
Cybersecurity research (vulnerability, threat analysis, malware, reverse engineering), LLM system development/evaluation, dataset/benchmark/grader design, post-training methods (SFT, RL, DPO), Python programming, system languages (C/C++/Rust/Java/Go), experimental design, failure analysis
Preferred skills
Reinforcement learning environments, DPO/RLHF/RLAIF, verifiable cybersecurity tasks, vulnerability research automation, distributed training, large-scale inference, research publications/patents
Technologies
Python, C, C++, Rust, Java, Go, containerized environments, testing frameworks, build systems, debuggers, malware sandboxes, fuzzing tools
Responsibilities
Design reproducible training and evaluation environments for complex cybersecurity tasks; Transform security data into structured tasks for LLMs; Develop high-quality security datasets (synthetic, adversarial, expert-annotated); Build reliable graders and reward signals; Design evaluations for security reasoning and autonomous task completion; Analyze model failures and capability regressions; Develop and evaluate post-training methods; Build iterative model-improvement loops; Conduct controlled experiments; Collaborate with security and ML teams to move research to production
Seniority
Principal, hands-on IC with strategy & mentorship