Jun Gu is the partner of Zilliz, performing the Senior Architect role. Before joining Zilliz, Jun received his undergraduate degree in Computer Science from Peking University and worked as a database technician for 14 years in companies like ICBC, IBM, Morgan Stanley, and Huawei. Jun is a PMC member of the Milvus project, an LF AI & Data incubation project. Jun has delivered several speeches on the Milvus project in different OSS summits hosted by the Linux Foundation.
How to accelerate approximate nearest neighborhood search (ANNS) for large scale datasets.
Introduction (Jun’s background, basic info on Zilliz) - 0:00
How to unlock the treasure of unstructured data (Jun describes the challenge of understanding and utilizing unstructured data) - 2:21
The flow-based AI applications (Jun talks about the most popular way of using AI technologies to analyze and structure data and the problem of data fragmentation that comes with it) - 4:03
The unstructured data service for AI (A more complex model for dealing with unstructured data and how Milvus fits into it) - 6:10
Vectors are different (Why traditional databases don’t meet the requirements of vector analysis and similarity comparison, segue into what Milvus does differently) - 11:12
Milvus: the big picture (A diagram describing different parts of Milvus project, support for different application development environments) - 13:26
The ANN benchmark (Jun talks about performance comparison and explains how Milvus manages to achieve faster search with lower memory consumption) - 16:08
Data management (Technical explanation of how Milvus does data management, what changed between different versions of Milvus) - 21:44
Our journey (A short story about Milvus and how it became one of the most active AI projects in Linux foundation, basic facts: number of users, releases, patents etc.) - 28:43
User scenarios (Jun walks us through three real world use cases: an intelligent writing assistant, image search for company trademark and pharmaceutical molecule analysis) - 30:07
Places to connect and read more about Milvus - 35:22
Share your questions and comments below!