Skytalk 2

Title: A New Era of Datacenter Scale AI Computing

Speaker: Lingjie Xu, Shanghai Biren Technology

Abstract: As modern deep neural networks have become more versatile and demanded more computing power, companies and researchers poured loads of resource into building high performance chips to address challenges of AI training and deployment in datacenter. Developers are keen to boost productivity and reduce overall TCO (total cost of ownership), without sacrificing programmability. Biren Technology is dedicated to providing world-leading computing capabilities and its flagship BR100 GPGPU chip brings best-in-class performance and efficiency, measured in both performance per watt and performance per dollar. This talk will discuss hardware and software co-design philosophy behind world’s fastest GPGPU chip and how BR100 empowers a new era of datacenter scale AI computing.

Bio: Lingjie Xu is the co-founder of Biren Technology, where he oversees product marketing and solution engineering. In the past decade, he held various management and architect roles in NVIDIA, AMD and Samsung, and helped bring many GPU products to market. Before cofounding Biren, he was a director at Alibaba Cloud, responsible for its heterogeneous computing infrastructure for AI applications. Lingjie was an early advocate of MLPerf community and later became the founding board member of MLCommons. He holds a B.S.E.E. degree from Shanghai Jiaotong University, an M.S.E.E degree from UT Austin, and an MBA degree from UC Berkeley.

About Biren Technology: Headquartered in Shanghai China, Biren Technology designs high performance accelerated computing chips and provides end-to-end platform solutions.