This RAG From Scratch course provides a complete step-by-step guide to understanding and building Retrieval-Augmented Generation (RAG) systems. It is designed for learners who want to move from basic LLM usage to advanced AI systems that can retrieve and use external knowledge effectively.
The course begins with an overview of RAG and explains how it improves large language models by combining them with external data sources. Learners then explore the full RAG pipeline starting with indexing, where documents are processed and stored in a searchable format.
Next, the course covers retrieval techniques, explaining how relevant information is selected from large datasets. It then moves into the generation phase, where retrieved data is used to produce accurate and context-aware responses.
A major part of the course focuses on query translation techniques, including multi-query, decomposition, step-back reasoning, and HyDE (Hypothetical Document Embeddings). These methods improve search accuracy and response quality in RAG systems.
The course also introduces routing and structured query processing, allowing systems to choose the best retrieval strategy dynamically. Advanced topics include multi-representation indexing and RAPTOR, which enhances hierarchical retrieval performance.
By the end of this course, learners will understand the complete RAG pipeline and be able to build advanced AI systems that combine retrieval, reasoning, and generation for real-world applications.
