Principal Security Researcher – AI (Cybersecurity LLM Post-Training, Evals, and Environments)

Our Mission At Palo Alto Networks®, we’re united by a shared mission—to protect our digital way of life. We thrive at the intersection of innovation and impact, solving real-world problems with cutting-edge technology and bold thinking. Here, everyone has a voice, and every idea counts. If you’re ready to do the most meaningful work of your career alongside people who are just as passionate as you are, you’re in the right place. Who We Are In order to be the cybersecurity partner of choice, we must trailblaze the path and shape the future of our industry. This is something our employees work at each day and is defined by our values: Disruption, Collaboration, Execution, Integrity, and Inclusion. We weave AI into the fabric of everything we do and use it to augment the impact every individual can have. If you are passionate about solving real-world problems and ideating beside the best and the brightest, we invite you to join us! We believe collaboration thrives in person. That’s why most of our teams work from the office full time, with flexibility when it’s needed. This model supports real-time problem-solving, stronger relationships, and the kind of precision that drives great outcomes. Job Summary Your Career As a Principal AI Researcher, you will advance the cybersecurity capabilities of large language models and autonomous AI agents by combining security research, rigorous evaluation, and applied LLM post-training. You will develop high-quality security data, realistic training and evaluation environments, reliable graders, and post-training methods for complex cybersecurity tasks spanning vulnerability research, threat analysis, detection, investigation, remediation, secure coding, and autonomous security workflows. You will work across security research, machine learning, and research engineering to identify model capability gaps and translate real-world cybersecurity problems into measurable model improvements. Your Impact Design reproducible training and evaluation environments for complex cybersecurity tasks, including vulnerability research, threat analysis, detection, investigation, remediation, secure coding, and autonomous security workflows. Transform source-code repositories, vulnerabilities, security incidents, malware samples, threat intelligence, detection logic, patches, test harnesses, and security tools into structured tasks for LLMs and AI agents. Develop high-quality security datasets, including synthetic data, hard negatives, adversarial examples, expert annotations, and model-generated trajectories. Build reliable graders, verifiers, and reward signals using tests, compilation, runtime behavior, security-tool outputs, detection results, vulnerability reproduction, and other domain-specific validation methods. Design evaluations for security reasoning, code understanding, threat analysis, root-cause analysis, tool use, long-horizon execution, and autonomous task completion. Analyze model failures, data quality issues, grader weaknesses, reward hacking, benchmark overfitting, and capability regressions. Develop and evaluate post-training methods, including supervised fine-tuning, preference optimization, reinforcement learning, reward modeling, rejection sampling, and distillation. Build iterative model-improvement loops using model rollouts, verifier feedback, expert review, synthetic data generation, and failure-driven data collection. Conduct controlled experiments with appropriate baselines, ablations, and regression testing. Collaborate with Security Researchers, ML Engineers, infrastructure teams, and product teams to move research into production capabilities. Qualifications Your Experience Required Qualifications: Strong hands-on experience in one or more cybersecurity areas, such as vulnerability research, secure coding, threat detection, malware analysis, incident investigation, reverse engineering, fuzzing, or security automation. Practical experience developing or evaluating LLM-based systems for cybersecurity, source code, software engineering, or tool-using agents. Experience designing datasets, benchmarks, graders, evaluation frameworks, or tasks with verifiable outcomes. Experience with at least one post-training or model-adaptation method, such as supervised fine-tuning, preference optimization, reinforcement learning, reward modeling, synthetic data generation, or distillation. Strong programming skills in Python and at least one system or application language, such as C, C++, Rust, Java, or Go. Experience working with large codebases, security tools, testing frameworks, build systems, debuggers, and containerized environments. Strong understanding of experimental design, failure analysis, regression testing, and quantitative evaluation. Ability to independently drive ambiguous research problems from investigation through implementation and validation. BS/MS degree in Computer Science, Machine Learning, Artificial Intelligence, Cybersecurity, or a related field, or equivalent practical experience. Preferred Qualifications: Experience building reinforcement-learning environments, cybersecurity agents, coding agents, or other tool-using AI systems. Experience with DPO, RLHF, RLAIF, online or offline reinforcement learning, process supervision, or model distillation. Experience developing verifiable cybersecurity tasks using tests, sanitizers, fuzzing, symbolic execution, malware sandboxes, detection systems, incident data, exploit reproduction, or patch validation. Experience with vulnerability research automation, AI-assisted security analysis, threat detection, malware analysis, incident response, static or dynamic analysis, or software supply-chain security. Experience training or evaluating code-focused or cybersecurity-focused language models and autonomous agents. Experience with distributed training, large-scale inference, rollout generation, or production ML platforms. Publications, open-source contributions, patents, vulnerability disclosures, or other demonstrated research impact in cybersecurity, LLM post-training, reinforcement learning, code intelligence, or AI agents. Compensation Disclosure The compensation offered for this position will depend on qualifications, experience, and work location. For candidates who receive an offer at the posted level, the starting base salary (for non-sales roles) or base salary + commission target (for sales/com-missioned roles) is expected to be the annual range listed below. The offered compensation may also include restricted stock units and a bonus. A description of our employee benefits may be found here . $163,200.00 - $264,000.00/yr Our Commitment We’re trailblazers that dream big, take risks, and challenge cybersecurity’s status quo. It’s simple: we can’t accomplish our mission without diverse teams innovating, together. We are committed to providing reasonable accommodations for all qualified individuals with a disability. If you require assistance or accommodation due to a disability or special need, please contact us at . Palo Alto Networks is an equal opportunity employer. We celebrate diversity in our workplace, and all qualified applicants will receive consideration for employment without regard to age, ancestry, color, family or medical care leave, gender identity or expression, genetic information, marital status, medical condition, national origin, physical or mental disability, political affiliation, protected veteran status, race, religion, sex (including pregnancy), sexual orientation, or other legally protected characteristics. All your information will be kept confidential according to EEO guidelines. Is role eligible for Immigration Sponsorship?: Yes