Yihao Cai

Institute of Computing Technology, Wayne State University

Papers

Total Citations

H-Index

About

Yihao Cai is at the forefront of efficient AI deployment and safety-critical machine learning systems. His research primarily focuses on bridging the gap between high-performance large language models (LLMs) and practical, real-world constraints—particularly in edge computing and cyber-physical systems (CPS). Cai’s most notable contribution is the "Cambricon-LLM," a chiplet-based hybrid architecture that enables on-device inference of massive 70-billion-parameter LLMs. This work, which has already garnered 17 citations since its 2024 publication, addresses the critical challenge of deploying advanced AI on resource-constrained devices like smartphones and robotics, preserving user privacy and network resilience without sacrificing intelligence. In his more recent 2025 work, Cai introduces the "Runtime Learning Machine," a novel framework for safety-critical CPS. This system features a tripartite architecture—a high-performance Student, a high-assurance Teacher, and a Coordinator—designed to maintain both performance and reliability in applications where failure is not an option. By tackling the tension between computational power and trustworthiness, Yihao Cai is shaping the future of deployable, dependable AI.