Talk Abstract:

The advent of large language models (LLMs) has transformed a wide spectrum of applications, from everyday tasks to complex business processes. This talk will take you on a journey from graph analytics to LLMs, starting with an overview of our recent advancements in graph analytics. The diverse application scenarios at Alibaba have driven the development of GraphScope Flex, a modular architecture designed for flexible deployment based on specific application needs. We will then explore how the integration of graph techniques with LLMs opens new possibilities for more intelligent and efficient data processing.  As part of the presentation, we will showcase a demo, “Graphy,” demonstrating how we use LLMs to convert unstructured data, such as PDF documents, into graph data to enable multi-document analysis. Lastly, we will provide an overview of our ongoing efforts to optimize LLM inference for improved efficiency and scalability, addressing the demands of real-world deployment.

Bio:

Dr. Longbin Lai received his Ph.D. from the University of New South Wales, Sydney, in 2017, where he continued as a postdoctoral researcher until 2019. He then joined Alibaba DAMO Academy, focusing on the development of distributed systems to support the vast scale of Alibaba's e-commerce platforms. Now, Dr. Lai becomes a member of Alibaba's Tongyi Lab, where he is working on developing large-scale inference engines to efficiently serve large language models and their applications within Alibaba and beyond.