Entrepreneurship Workshop
"TBC"
Dr. Yan ZHENG (Riverhead Capital Investment Management Co., Ltd)
Abstract
TBC
Dr. Yan ZHENG (Riverhead Capital Investment Management Co., Ltd)
"Introduction and case study on ZJU-IC Industry & Academic Collaboration"
Dr. Feijun ZHENG (Zhejiang University)
Abstract
The principles of IC college at Zhejiang University is integration of industry and R&D, collaboration of science and education, we are actively exploring a new type of education and scientific research system and system innovation oriented to industry needs. With the 300mm 55nm CMOS mini-line based-in Zhejiang University, Hangzhou, we bring together brilliant ideas from university to support IC Industry. We provide Industry-level facilities and capability for R&D and low-volume manufacturing. With the capability of DTCO, providing standard and customized process development for customers. The institute of VLSI Design focus on embedded IP core and processor design field, the series of CPU cores have accumulated more than 1 billion devices.
Dr. Feijun ZHENG (Zhejiang University)
"Building Computing Systems for Embodied Artificial Intelligence"
Dr. Shaoshan LIU (Shenzhen Institute of Artificial Intelligence and Robotics for Society (AIRS))
Abstract
Embodied AI integrates artificial intelligence into physical systems, enabling robots to perceive, learn, and interact with dynamic environments. However, its development faces key computing challenges: integrating diverse functions like perception and control into cohesive software, processing multimodal sensor data in real time under energy and concurrency constraints, and overcoming data scarcity due to the difficulty of collecting diverse real-world robotic interactions. To address these issues, a layered software architecture is proposed to abstract hardware complexity, streamline function integration, and support foundational AI models. A novel computing architecture enhances efficiency through synchronized sensor fusion, dataflow accelerators, and hardware-software co-design, enabling real-time AI execution on edge devices. Additionally, design automation leverages synthetic data and digital twin simulations to train models in virtual environments, minimizing reliance on extensive physical data collection. Together, these approaches aim to enhance the flexibility, efficiency, and scalability of embodied AI systems.
Dr. Shaoshan LIU (Shenzhen Institute of Artificial Intelligence and Robotics for Society (AIRS))