Build the hardest parts personally: prototype agent loops, tune retrieval, design evals, and ship the shared infrastructure that powers AI Case Explorer (case overviews, timelines, key people and themes, PII surfacing, and Agent chat), AI Agent Review (pre-classifying relevance, privilege, and key issues, shipping MLP), and CS & OPS tech-enablement using AI as a core part of our central work orchestration system. Define and implement how we select, triage, and route across models (Azure OpenAI, Anthropic, open-source, fine-tuned), manage vector databases and retrieval, and evolve our agent harness as the frontier moves.