For engineers who can call an LLM API and now need to ground it in their organization's documents: design, build, evaluate and operate a RAG system that holds up in production.
Level
Intermediate
Study time
14 hours
Lessons
16 in 6 modules
Exam
20 questions
About this course
Most RAG prototypes work on the demo questions and fail quietly on everything else. This course teaches the full pipeline as an engineering system with measurable parts: parsing and chunking that preserve structure, embedding model selection, approximate nearest-neighbor indexes such as HNSW, hybrid lexical and dense retrieval fused with reciprocal rank fusion, cross-encoder reranking, query rewriting, and prompt assembly that produces verifiable citations.
The second half covers what separates a prototype from a system people rely on: enforcing document permissions at retrieval time, building an evaluation set and tracking recall@k, MRR and faithfulness, a stage-by-stage debugging method, agentic retrieval loops, and the operations work of re-indexing, freshness and cost control. Code is vendor-neutral Python with notes on pgvector, FAISS, Elasticsearch/OpenSearch and Claude's search-result citations where they apply. The course ends in an automated exam; passing earns a Wirewalk Institute of AI certificate of completion and assessment.
What you will be able to do
Decide between RAG, fine-tuning and long-context prompting for a given use case and justify the choice on freshness, scale, access control and cost
Build an ingestion pipeline that parses, cleans and chunks documents while preserving structure and metadata needed for citations and permissions
Select an embedding model and configure a vector index such as HNSW, explaining the recall, latency and memory trade-offs of its parameters
Implement hybrid retrieval with BM25 and dense vectors fused by reciprocal rank fusion, followed by cross-encoder reranking and query rewriting
Assemble grounded prompts that produce checkable citations and enforce document-level permissions before any content reaches the model
Build a labeled evaluation set and measure recall@k, MRR and faithfulness to compare pipeline changes objectively
Diagnose RAG failures by isolating the failing stage, and operate the system through re-indexing, freshness monitoring and cost control
Syllabus
6 modules · 16 lessons · about 14 hours
1
Module 1: RAG Foundations and Architecture
When retrieval is the right tool, and the anatomy of a minimal end-to-end pipeline.2 lessons
RAG vs Fine-Tuning vs Long Context: Choosing the Right Tool
Anatomy of a RAG Pipeline: A Minimal Working System
2
Module 2: Ingestion, Chunking and Embeddings
Turning messy source documents into well-formed, well-embedded retrieval units.3 lessons
Ingestion and Parsing: PDFs, HTML, Tables and Metadata
Chunking Strategies That Preserve Meaning
Embeddings: How They Work and How to Choose a Model
3
Module 3: Retrieval: Indexes, Hybrid Search and Reranking
Finding the right chunks quickly: ANN indexes, lexical plus dense search, reranking and query rewriting.4 lessons
Vector Indexes: Flat, IVF and HNSW in Practice
Hybrid Search: Combining BM25 and Vectors with Rank Fusion
Reranking with Cross-Encoders
Query Rewriting, Expansion and Decomposition
4
Module 4: Generation and Access Control
Assembling grounded prompts with citations, and making sure users only retrieve what they may see.2 lessons
Prompt Assembly and Grounded Citations
Permission-Aware Retrieval
5
Module 5: Evaluation and Debugging
Measuring retrieval and answer quality, and finding which stage is actually broken.3 lessons
Retrieval Metrics: Recall@k, MRR and Building a Golden Set
Evaluating Answers: Faithfulness, Relevance and LLM Judges
RAG Failure Modes and a Stage-by-Stage Debugging Method
6
Module 6: Agentic Retrieval and Operations
Letting the model drive retrieval, and running the system day to day.2 lessons
Agentic Retrieval: Retrieval as a Tool
Operating RAG: Re-Indexing, Freshness and Cost
Exam and certificate
The exam unlocks once you have completed every lesson. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Anthropic, PostgreSQL pgvector, FAISS, Elasticsearch, OpenSearch, or Hugging Face Sentence Transformers. Product and company names are trademarks of their respective owners.
A hands-on course on building no-code assistants with the builders already in your workplace tools: ChatGPT projects, custom GPTs and their successor plugins, Claude Projects and Skills, Gemini Gems, and Microsoft 365 Copilot Agent Builder and Copilot Studio. You leave with a tested, documented assistant and a plan to share it safely.
A hands-on, code-first course on the three major LLM APIs side by side: messages, streaming, structured outputs, tool calling, vision, cost, caching, batch, retries, abstraction layers, cloud routes and testing. You finish by building and hardening a small multi-vendor service.
A deep dive into the two layers under every semantic search system: embedding models and vector databases. Covers how embeddings are trained and benchmarked, Matryoshka truncation and quantization, ANN index internals (HNSW, IVF-PQ, DiskANN), filtering, multi-tenancy and the operational work of re-embedding and capacity planning.