Research Scientist (AI Behaviours)

White Circle is seeking a Research Scientist to study how LLM agents fail in the wild, focusing on deception, misalignment, and unsafe behavior. The role involves designing experiments, building automated audit agents, and publishing findings. The ideal candidate has a strong background in ML engineering and empirical research.

• Own research projects end to end, from defining the question to running experiments and publishing results.
• Develop automated audit agents that discover and characterize suspect model behavior at scale.
• Study how misalignment and bias appear in real user interactions with agents.
• Pressure-test frontier agents in realistic, high-stakes scenarios.
• Run white-box and black-box investigations to understand how AI models fail.
• Publish findings as blog posts and conference papers.

• Track record of empirical research in agent behavior, model evaluation, alignment, or adjacent area.
• Strong ML engineering skills to independently build research MVPs involving fine-tuning, agent inference, and evals.
• Skills in experimental design: isolating failure modes, calibrating judges, distinguishing signal from artifact.
• Ability to define experiments from vague behavioral questions and iterate quickly.