Purdue CS592: Databases for AI: Vector Databases
(Spring 2026)

Course Description

Vector databases have recently emerged as a hot topic in the broader realm of databases for AI. The surge of interest is largely fueled by large language models (LLMs), where vector databases help overcome inherent limitations such as hallucinations, lack of domain expertise, and the inability to incorporate real-time information. This is enabled by the new paradigm of Retrieval-Augmented Generation (RAG), in which vector databases act as external knowledge bases, delivering relevant context to LLMs via vector search. While vector search itself is not new, modern vector databases face a host of system-level challenges, which we will explore in depth in this course.

In this seminar, we will first cover the fundamentals of vector databases. Then we will feature a series of invited talks and student presentations covering recent advances in vector databases. By the end of the course, students are expected to gain a solid understanding of the challenges, the state-of-the-art techniques, and the open problems in vector databases.

Instructor

• Jianguo Wang
• Email: csjgwang@purdue.edu

Logistics

• When: MW 4:30-5:45pm
• Where: GRIS 133
• Office hour: after class or make appointment
• Pre-requisites: No prior experience with vector databases is required. However, familiarity with data structures (e.g., CS251), databases (e.g., CS348 or CS448), and introductory AI/ML (e.g., CS242 or CS243) will be helpful.

Online communications

• We'll use Piazza, e.g., announcements, discussions, and Q&A.
• We'll NOT use Brightspace except for sending emails occasionally.

Schedule (More Speakers to Be Invited)

Lecture

Talk/Paper Title

Presenter