Abstract: 

This workshop will show you how a vector database can help to scale the power of modern deep-learning models and effortlessly combine them with your data. More specifically, you will learn about how vector databases can help you to work with large-scale vector embeddings, and integrate the power of large language models (LLMs).

By the end of the workshop, you will have a good overview of what a vector database is, and be able to perform multiple types of searches yourself, including similarity searches, hybrid searches and generative searches.

You will learn how vector databases work, and get hands-on with one yourself. You will learn to perform similarity searches using state-of-the-art models and across multiple languages. You will also learn to perform generative searches, where your data is transformed with LLMs at retrieval time, to bring your data to life. Throughout this workshop, you will see how vector databases enable users to leverage the power of LLMs and apply them to their own data at scale at lightning-fast speeds.

We will use Weaviate (an open-source vector database) and its Python client to run this workshop.

This workshop will begin with a demonstration of capabilities before getting onto a hands-on portion. It will also briefly cover core concepts such as what vector embeddings are, how vector search works, and what LLMs do. You will also have opportunities to ask questions and discuss with participants.

The workshop will be run with Jupyter Notebooks in an online environment, so if you have a computer, intermediate-level Python skills and and a keen interest in AI, we would love to see you there.

Session Outline:

Module 1: An introduction to vector databases
- You will see some of the things that vector databases can do, before learning about what they are and how they work. By the end, you will be able to broadly describe what vector databases are and their capabilities.

Module 2: Hands on with a vector database
- You will start to run queries yourself on an existing vector database. You'll be able to try out various types of queries including generative searches that leverage the power of large language models. By the end, you will be able to run various queries with Weaviate.

Module 3: Build your own vector database
- Continuing on from module 2, you will learn how to populate a vector database instance with Weaviate. You'll be spinning up your own instance of Weaviate, downloading a real dataset, and populating a vector database with it. By the the end, you will be able to build a vector database.

Module 4: Making a vector database work for you / Q&A
- Here, you will learn about various details that will help to make a vector database work well for you. We'll discuss how to think about vectors, indexing, and schema design, so that you can build not just a vector database, but *your* vector database.

Background Knowledge:

A working knowledge of Python is sufficient, as well as a keen interest in AI or language models. Experience with Jupyter Notebooks will help to follow along, but is not required.

Bio: 

JP finds joy in technology and learning, as well as empowering others by helping to distill complex technologies into relatable concepts. He works at Weaviate as the Technical Curriculum Coordinator, facilitating education for vector databases and data science topics. When he’s not working, JP enjoys immersing himself in the worlds of games and sports. You might spot him working on his serve on the tennis court, or engaging in spirited board game sessions.