A model only knows what it was trained on, and it knows nothing about your company documents, your product data, or anything newer than its training. Retrieval-Augmented Generation, or RAG, solves this by finding the relevant information first and placing it in the prompt, so the model answers from your data instead of guessing.
It starts with how RAG works end to end: loading documents, turning text into embeddings, and searching a vector database. It then looks at chunking, the choice of how to split documents, which has a large effect on answer quality, and finishes with techniques for improving retrieval, such as adding context to chunks and reranking results.
Watch the lessons in order, then continue on to Agents & Workflows.
Why plain LLMs hallucinate and go out of date, then the full RAG pipeline: ingestion, chunking, embeddings, vector databases, similarity search, and generating an answer from the retrieved context.
How a vector database fits into a RAG application, and what it replaces compared with searching and managing vectors by hand.
Why chunking matters in a RAG system and the main strategies for splitting documents before indexing.
A retrieval technique that combines chunking with reranking to improve how well a RAG system finds the right passages.