Abstract: 

The rapid advancement of AI has led to the emergence of massive, complex foundational models that require enormous computational resources, making efficient training and inference systems essential. Training these models involves distributing large-scale computations, optimizing resource usage through computation-communication overlap, employing advanced parallelization strategies, and ensuring fault tolerance at scale. Inference systems, on the other hand, must handle diverse workloads with varying SLAs, quickly adopt engineering optimizations, and balance trade-offs between throughput, latency, cost, and availability, particularly in distributed settings. In this talk, I will explore the key systems challenges in developing foundational models, including Qwen (large language models) and Wan (video generative models), and present our ongoing approaches to both training and inference at Alibaba. These designs significantly improve the efficiency of managing complex AI workloads in cloud environments.

Bio:

Jingren Zhou is the Chief Technology Officer at Alibaba Cloud, where he drives technology innovation and product development across a broad range of cloud computing services. He also leads the development of foundational AI models, including the Qwen and Wan models, and their applications in diverse business applications within Alibaba Cloud. Prior to this role, he played a key role in building Alibaba’s cloud-scale distributed data analytics platform and developing advanced techniques for personalized search, product recommendation, and advertising on Alibaba’s e-commerce platform. Before joining Alibaba, he was a veteran at Microsoft, focusing on big data and database research and development. His research interests include cloud computing, distributed systems, databases, and large-scale machine learning. He has served as PC co-chair and core committee member for many academic conferences and technical forums. He received his PhD in Computer Science from Columbia University. He is a Fellow of ACM and IEEE.

Abstract: 

Advances in AI are making it increasingly practical to extend traditional database management systems with the ability to query unstructured data, including images and videos, thus producing multimodal database management systems. In this talk, we will discuss some of the key components required for this integration, and will then focus specifically on video data management. A key challenge with video data management is the ability to cost-effectively

support complex queries in diverse applications and over video data from a variety of domains. We will present the VOCAL system and how it uses vision language models (VLMs) to achieve self-enhancing video data management, where the system extends its functionality to cost-effectively support compositional queries expressed in natural language in a variety of domains, increasing its capabilities with usage. 

Bio:

Magdalena Balazinska is Professor, Bill & Melinda Gates Chair, and Director of the Paul G. Allen School of Computer Science & Engineering at the University of Washington. Her research focuses on database management systems with a current focus on video data management, multimodal data management, and generally connections between AI and data management. Prior to her leadership of the Allen School, Magdalena was the Director of the eScience Institute and the Associate Vice Provost for Data Science. Magdalena is an ACM Fellow. She holds a Ph.D. from the Massachusetts Institute of Technology (2006). Shortly after her arrival at the University of Washington, she was named a Microsoft Research New Faculty Fellow (2007). She also received the inaugural VLDB Women in Database Research Award (2016) for her work on scalable distributed data systems, and both a CIDR Test-of-Time Award (2025) and the ACM SIGMOD Test-of-Time Award (2017) for her work on stream processing systems, a 10-year most influential paper award (2010) from her earlier work on reengineering software clones, and other best-paper and "best of" awards. 

Talk Slides: MIDAS-AcademicKeynote-Balazinska-Magdalena Balazinska-20250621-22.25.pptx