05/27/2026
Tomorrow I’ll be teaching Retrieval-Augmented Generation as part of an AI certificate program at Saint Louis University.
RAG is one of the most important practical patterns in applied AI right now because it helps move large language models beyond general-purpose responses and into domain-specific, evidence-supported answers.
At a high level, RAG connects an AI model to external knowledge sources such as documents, policies, manuals, databases, knowledge bases, or other structured and unstructured data. Instead of relying solely on what the model learned during training, the system retrieves relevant information at the time of the user’s question and uses it to generate a more grounded response.
That matters because many real-world AI use cases depend on current, accurate, and organization-specific information. A cybersecurity analyst may need to query internal procedures. A compliance team may need to search policies, control mappings, or audit evidence. A support team may need answers based on product documentation. In each case, the value of the AI system depends less on “chatbot magic” and more on whether it can retrieve the right information, use it correctly, and explain the result clearly.
In this course, we’ll look at the key components of that workflow, including embeddings, vector databases, retrieval strategies, prompt augmentation, grounding, citations, evaluation, and the design of RAG pipelines. We’ll also connect those concepts to real-world AI workflows where systems need to retrieve information, use tools, and support more context-aware decision-making.
This is one of the areas where AI becomes much more practical for business, cybersecurity, compliance, and operations. It is also where good architecture matters. The model is only one part of the system. The data, retrieval method, context window, prompt design, evaluation process, and governance controls all matter.
I’m looking forward to helping students break down these concepts and see how RAG can be used to build more useful, reliable, and grounded AI systems.
