Intermediate·1 hour·6 lessons

RAG and Knowledge Systems

Build applications that answer questions over private data — starting with the architectural decision most tutorials skip: do you actually need RAG, or is long context + prompt caching the better path?

What you'll learn

When RAG vs. long context

Document chunking strategies

Embeddings & vector databases

Retrieval quality & re-ranking

Combining RAG with tool use

Production RAG patterns

Your instructor

Priya Rajan

Staff Engineer, AI Platform, Anthropic

Priya has designed and scaled AI infrastructure.

Syllabus

Do You Actually Need RAG?

Evaluate whether your application needs a retrieval pipeline or if Claude's long context window and prompt caching are the better path.

Long Context Done Right

Structure documents for maximum retrieval quality, slash costs with prompt caching, and get machine-verifiable source attribution with the Citations API.

Building a Retrieval Pipeline

Build a production retrieval pipeline from chunking through contextual retrieval, embedding, hybrid search, and reranking.

Citations, Search Results, and Grounded Answers

Feed retrieved chunks to Claude using search result content blocks and custom documents, then handle retrieval failures so your system stays grounded.

MCP Servers for Knowledge Systems

Connect retrieval pipelines to any Claude-powered application using the Model Context Protocol, turning your knowledge base into a universal service.

Production Patterns and Cost Engineering

Optimize your knowledge system for cost, quality, and freshness, then use a decision matrix to choose the right architecture for your use case.

This course includes

6 self-paced lessons
1 hour of content
Claude tutor on every lesson
Certificate of completion

Free to start. No credit card required.