Build and run the AI evaluation engine: define what "correct" means per attribute, curate golden sets with the research team, stand up LLM-as-judge gates, and set confidence thresholds for auto-accept, auto-reject, and human adjudication. Own Person Data end to end, setting the strategy, roadmap, and monthly priorities for contact accuracy, coverage, freshness, and compliance, from cleaning up bogus executive contacts to leadership extraction and person location.