Zhiyu Li

Co-Founder & CTO, MemTensor

Zhiyu Li, Ph.D., is an Oriental Talent Program Top Talent in Shanghai and a 36Kr “36Under36” Science and Technology Innovator. He is Co-Founder and CTO of Memory Tensor (Shanghai) Technology Co., Ltd., Technical Lead and Researcher of the Large Model Center at the Shanghai Institute of Algorithm Innovation. He has long been engaged in R&D and technical innovation in pre-training and large model applications, with research interests including memory augmentation for large models, efficient evaluation, and application algorithms. Previously, Li led multiple core algorithm initiatives at leading technology companies including Alibaba and Xiaohongshu. His technical achievements have been applied to large-scale business scenarios such as product reviews, Double 11 shopping festivals, and marketing advertising, generating cumulative revenue of several billion RMB and impacting nearly 100 million users. He also received the Double 11 Technology Breakthrough Award. In 2024, Li and his team proposed the first large model with an innovative hierarchical memory architecture, which attracted widespread attention in the industry. In 2025, he proposed the industry's first large model memory operating system, MemOS, which has accumulated more than 11K+ GitHub Stars and over 20,000 developers. The project subsequently received multiple awards, including the 2025 AI Infrastructure Excellence Award and the 2025 AI Technology Innovation Breakthrough Award. In 2026, he proposed Metis 1.0, the first memory-native foundation model, and introduced a definition and foundational theory of native memory. These large model technologies have been deployed in multiple state-owned enterprises and central government-owned enterprises, including Bank of China, China Merchants Securities, China Telecom, and Xinhua News Agency. To date, Li has published more than 70 papers in international conferences and journals including Patterns (Cell Press), NeurIPS, ICLR, ACL, and TKDE, and holds more than 10 granted patents. He is also the author of Memory Engineering: Building and Practicing Agent Memory Systems, the first professional book in the field of AI memory. He currently serves as a committee member of the Information Retrieval Committee and the Large Models and Generation Committee of the Chinese Information Processing Society of China. His research has been featured on the cover of MIT Technology Review and has repeatedly ranked No. 1 among trending papers on Hugging Face.