Lead Rigorous Model Evaluations: Architect and execute comprehensive evaluation suites for LLMs and multimodal models, identifying edge cases in multi-step reasoning, factuality, adversarial robustness, safety, and alignment. Advanced Scoring Frameworks: Develop deterministic, heuristic, and LLM-assisted evaluation frameworks (e.g., LLM-as-a-judge, reward modeling) to quantify human-perceived quality metrics (e.g., helpfulness, hallucination rates).