Tzu-Hao Liu is a Research Pathway student focused on efficient inference for large language models (LLMs). His research emphasizes accelerating the decoding stage through speculative decoding (SD) and related optimization techniques, contributing to the development of faster and more efficient LLM systems.