Liang Du works across foundation models and data‑centric AI—scaling and conversion of LLMs, principled data curation, and agent evaluation. He develops executable knowledge graphs for reproducible research and advances automated reasoning and reinforcement learning for real‑world operations.