Research Scientist/Engineer (Agentic Systems)

Responsibilities: Study how agentic systems plan, use tools, and behave over time, then turn those insights into practical runtime security controls.

• Research failure modes in agents, tool-calling workflows, memory, planning, and multi-step execution.
• Design experiments and benchmarks for off-task behavior, excessive agency, manipulation, and unsafe actions.
• Build prototypes, datasets, and evaluation pipelines to test hypotheses quickly.
• Translate research into policies, detections, enforcement logic, and product requirements.
• Track new agent architectures and attack techniques and assess their real-world impact.
• Communicate results through clear reports, demos, and close collaboration with engineering and product.

Ideal candidate: You have research or engineering experience with agents, LLMs, reinforcement learning, planning, systems, or AI security. You are comfortable forming hypotheses, building experiments, and shipping code. You can reason about multi-step behavior and distinguish interesting academic failure modes from material customer risk. Strong Python, clear writing, and practical curiosity are essential; publications or open-source work are a plus.