RAG Systems courses

• Understand Retrieval-Augmented Generation (RAG) and how retrieval can improve knowledge-intensive large language model applications by grounding generation in relevant external information.
• Explore the complete RAG pipeline from knowledge ingestion, indexing, and retrieval through context augmentation and final response generation.
• Compare RAG with prompting, fine-tuning, and long-context approaches, and identify the situations in which each strategy or combination of strategies is most appropriate.
• Design foundational RAG architectures for real-world applications by connecting knowledge sources, retrieval components, context construction, and language-model generation.

• Build ingestion pipelines for PDFs, webpages, databases, office documents, and other enterprise knowledge sources that must be made available to RAG applications.
• Apply parsing, cleaning, chunking, metadata enrichment, document transformation, and normalization strategies to prepare heterogeneous information for retrieval.
• Understand how document structure, semantic boundaries, overlap, chunk size, and metadata choices influence retrieval relevance and downstream answer quality.
• Prepare reliable and scalable knowledge bases with repeatable ingestion processes that can support evolving organizational information and production RAG systems.

• Understand embeddings, similarity measures, vector representations, and the principles that enable semantic retrieval based on meaning rather than exact keyword matching.
• Create, store, index, update, and query embeddings using vector databases and appropriate metadata structures.
• Compare dense, sparse, semantic, keyword, and hybrid retrieval approaches and understand the advantages and limitations of each method.
• Optimize indexing and search strategies for relevance, retrieval speed, scalability, infrastructure requirements, and operational cost.

• Design effective retrieval pipelines using query rewriting, expansion, decomposition, filtering, routing, and other techniques that improve the quality of information retrieved for a user request.
• Apply metadata filtering, hybrid search, reranking, and multi-stage retrieval to improve precision and surface the most relevant evidence.
• Handle ambiguous, complex, multi-part, and domain-specific queries by transforming them into retrieval operations that better reflect the user's information need.
• Diagnose retrieval failures systematically and improve document relevance, ranking, recall, and coverage using evaluation evidence.

• Transform retrieved information into high-quality context that large language models can interpret effectively when generating grounded responses.
• Apply context selection, compression, ordering, deduplication, grouping, and token-budget management to maximize the usefulness of retrieved evidence.
• Design prompts and context structures that encourage grounded generation, appropriate source usage, transparent uncertainty, and adherence to retrieved information.
• Balance context completeness and relevance against context-window limits, generation latency, model performance, and operational cost.

• Move beyond basic retrieve-and-generate pipelines by exploring adaptive, corrective, iterative, multi-step, and other advanced RAG architectures.
• Build systems that dynamically determine whether retrieval is required and decide what information to retrieve, where to retrieve it from, and how retrieval should evolve during task execution.
• Combine retrieval with tools, agents, planning, memory, state management, and iterative reasoning workflows to address more complex information needs.
• Select and design advanced RAG architectures according to application complexity, knowledge characteristics, reliability requirements, and expected user workflows.

See also: AI Agents courses →

• Extend RAG beyond plain text by retrieving and reasoning over tables, images, diagrams, structured records, and other multimodal document elements.
• Combine vector retrieval with knowledge graphs, relational databases, structured repositories, and other information sources to support richer knowledge access.
• Explore Graph RAG approaches for representing entities, relationships, connected knowledge, communities, and complex information structures that are difficult to capture through isolated chunks.
• Design hybrid retrieval architectures that combine multiple knowledge representations and retrieval methods to answer complex questions across heterogeneous organizational information.

• Evaluate retrieval quality using relevance, recall, precision, ranking, coverage, and context-quality measures that reflect the information needs of the target application.
• Measure generation quality through groundedness, faithfulness, factual correctness, completeness, relevance, citation accuracy, and other application-specific criteria.
• Build representative evaluation datasets, test cases, automated evaluation pipelines, and regression tests that support systematic RAG quality assurance.
• Diagnose whether application failures originate from ingestion, the knowledge base, retrieval, ranking, context construction, prompting, or final generation so that improvements target the correct component.

• Understand prompt injection through retrieved content, malicious documents, data leakage, retrieval poisoning, unauthorized knowledge access, and other security threats affecting RAG systems.
• Implement document permissions, identity-aware retrieval, metadata-based access controls, filtering, authorization checks, and secure retrieval mechanisms.
• Protect sensitive enterprise knowledge throughout ingestion, storage, indexing, retrieval, context construction, generation, logging, and monitoring.
• Build traceable and governed RAG systems with appropriate access control, auditability, source tracking, security monitoring, and human oversight for security-sensitive environments.

See also: AI Security courses →

• Architect end-to-end production RAG systems that integrate data ingestion, processing, indexing, retrieval, context construction, generation, evaluation, monitoring, and application interfaces.
• Optimize latency, scalability, caching, concurrency, reliability, infrastructure utilization, model usage, and operational cost across the complete RAG pipeline.
• Manage knowledge updates, incremental ingestion, re-indexing, versioning, observability, evaluation, incident investigation, and continuous quality improvement as organizational information changes.
• Build and deploy a production-ready RAG application designed to evolve safely and reliably with changing knowledge sources, application requirements, and user needs.