Postdoctoral Associate – AI Security

About the role

The Applied Research Laboratory for Intelligence & Security (ARLIS) at the University of Maryland seeks a Postdoctoral Associate to advance AI security research. You will work on cutting‑edge projects that protect advanced AI systems from sophisticated adversaries, supporting government and Intelligence Community partners.

Key responsibilities

• Conduct original research in AI security, including adversarial machine learning and model robustness.
• Develop and evaluate novel attack and defense techniques for modern AI systems, such as mechanistic and white‑box analysis, multi‑turn adaptive adversarial interactions, and security of reasoning and agent‑based architectures.
• Design experimental frameworks to assess AI system vulnerabilities across deployment scenarios (open‑weight, API‑based, hybrid).
• Apply interpretability methods—circuit analysis, feature attribution, sparse autoencoders—to uncover internal model behavior and failure modes.
• Contribute to benchmarks, evaluation methodologies, and datasets for AI security research.
• Collaborate with interdisciplinary teams to translate research findings into operational capabilities.

Required profile

• Post‑doctoral level researcher with a strong background in machine learning, cybersecurity, or related fields.

Required skills

• Adversarial machine learning
• Model robustness and secure AI system design
• Mechanistic and white‑box model analysis
• Multi‑turn adaptive adversarial interaction techniques
• Security of reasoning models and agentic architectures
• Interpretability techniques (circuit analysis, feature attribution, sparse autoencoders)

What we offer

• Opportunity to publish in top‑tier venues and influence national security research.
• Access to a mission‑driven R&D environment with interdisciplinary collaboration.
• Professional growth within a culture that values integrity and collaboration.