Conduct research on AI evaluation, rubrics, benchmarks, human judgment, preference data, and agent performance. Design rigorous experiments, validate measurement frameworks, build evaluation pipelines, and create long-horizon real-user benchmarks. Collaborate with engineers, delivery teams, customers, academics, and external labs, while publishing methods and benchmark findings. The role combines applied statistics, messy human data, research prototyping, technical writing, and deployment of research methods in live customer programs.