Generative AI

AIW02 Vectorization Fundamentals (Including Database Support)


8:00am - 9:15am

Level: Intermediate

Lino Tadros

President & CEO

The Training Boss LLC

In the world of Generative AI, passing a whole document, bunch of documents, database with multiple tables, datalakes or other source of knowledge bases to augment the prompt for LLM (RAG) requires what we call Embedding and Vectorization to reduce the amount of token usage during the LLM call but most importantly to perform more accurate retrieval based on semantic meaning. In this session, you will learn about the process of ingesting data, extracting data (Chunking and overlapping), embedding (using ada-002 and ada-003) and vectorization of the data and the chat question asked to retrieve accurate and meaningful response. We will also discuss multiple Vector databases that you can use to store the vectorization results like Pinecone, FAISS, ChromaDB, Azure PostgresSQL and Azure AI Search.