Research Engineer, AI Security & Defensive ML

Techeurope; Zürich; Hybrid; CHF 332,000 - 397,000; Full time

Job summary

Anthropic is seeking a Research Engineer on the Security team to safely advance AI model capabilities through secure coding, vulnerability remediation, and other defensive cybersecurity research. This role blends research and engineering, designing RL environments, running experiments, and delivering results into production training runs.

You will collaborate with researchers, engineers, and cybersecurity experts, both inside and outside Anthropic, to strengthen defensive workflows.

Qualifications

• Domain expertise in cybersecurity paired with interest or experience in training safe AI models.
• Experience designing and implementing RL environments and evaluating model capabilities.

Responsibilities

• Design, build, and maintain RL environments and evaluations that measure and improve model capabilities in secure coding, vulnerability discovery and remediation, and other defensive security tasks
• Run experiments to develop and test new training approaches, and deliver successful results into production training runs
• Partner with researchers, engineers, and security specialists across Anthropic to understand where AI can strengthen defensive workflows, and build toward those needs
• Work with external cybersecurity experts and organizations to ground environments and evaluations in realistic defensive work
• Contribute to the strategic direction of the team, including decisions about what to build, what to partner on, and where to invest

Job description

Anthropic is seeking a Research Engineer on the Security team to safely advance AI model capabilities through secure coding, vulnerability remediation, and other defensive cybersecurity research. This role blends research and engineering, designing RL environments, running experiments, and delivering results into production training runs.

You will collaborate with researchers, engineers, and cybersecurity experts, both inside and outside Anthropic, to strengthen defensive workflows.