Module 0: Start Here
Welcome to the Course
How each module works
Setting Up the Code
How to Read the Doodles
The visual language
The text walkthrough
Reuse them
A worked example
Finishing the Course
Module 1: The Knowledge Map: Source Contracts and Access Paths
Watch: The Knowledge Map: Source Contracts and Access Paths
Read: The Knowledge Map: Source Contracts and Access Paths
The Knowledge Sources Map
Retrieve or Look Up?
Tables vs Text
Sidebar: RAG vs Long Context
Further Reading
Next
Doodle Cheat Sheet: The Knowledge Map: Source Contracts and Access Paths
Cheat Sheet at a Glance
Exercise: Route the stock question
Check Your Work
Quiz: The Knowledge Map
99 attempts allowed
Module 2: Ingestion Pipelines: Stable IDs, Provenance and Re-Runs
Watch: Ingestion Pipelines: Stable IDs, Provenance and Re-Runs
Read: Ingestion Pipelines: Stable IDs, Provenance and Re-Runs
The Ingestion Conveyor
Every Chunk Knows Its Source
Further Reading
Next
Doodle Cheat Sheet: Ingestion Pipelines: Stable IDs, Provenance and Re-Runs
Cheat Sheet at a Glance
Exercise: Fail halfway
Check Your Work
Quiz: Ingestion Pipelines
99 attempts allowed
Module 3: Documents With Layout: Tables, Scans and Images
Watch: Documents With Layout: Tables, Scans and Images
Read: Documents With Layout: Tables, Scans and Images
The Messy PDF
Read the Page, Not Just the Text
Sidebar: OCR vs Native Text Extraction
Further Reading
Next
Doodle Cheat Sheet: Documents With Layout: Tables, Scans and Images
Cheat Sheet at a Glance
Exercise: Add the Swiss row
Check Your Work
Quiz: Documents With Layout
99 attempts allowed
Module 4: Structure-Aware Chunking and Applicability Metadata
Watch: Structure-Aware Chunking and Applicability Metadata
Read: Structure-Aware Chunking and Applicability Metadata
Chunks That Follow the Document
Small to Find, Big to Read
Metadata Tags and Filters
Sidebar: The Overlap Trade-off
Further Reading
Next
Doodle Cheat Sheet: Structure-Aware Chunking and Applicability Metadata
Cheat Sheet at a Glance
Exercise: The rule that starts next month
Check Your Work
Quiz: Structure-Aware Chunking and Applicability Metadata
99 attempts allowed
Module 5: Knowledge Access: Permissions, Tenants and Poisoned Documents
Watch: Knowledge Access: Permissions, Tenants and Poisoned Documents
Read: Knowledge Access: Permissions, Tenants and Poisoned Documents
Who May See This?
Poisoned Pages
Further Reading
Next
Doodle Cheat Sheet: Knowledge Access: Permissions, Tenants and Poisoned Documents
Cheat Sheet at a Glance
Exercise: Revoke and re-ask
Check Your Work
Quiz: Knowledge Access
99 attempts allowed
Module 6: Embedding Models and Vector Indexes
Watch: Embedding Models and Vector Indexes
Read: Embedding Models and Vector Indexes
Choosing an Embedding Model
The Approximate Neighbourhood
Sidebar: Flat, Graph and Clustered Indexes
Further Reading
Next
Doodle Cheat Sheet: Embedding Models and Vector Indexes
Cheat Sheet at a Glance
Exercise: Search one more cluster
Check Your Work
Quiz: Embedding Models and Vector Indexes
99 attempts allowed
Module 7: Hybrid Search: Keywords Plus Meaning
Watch: Hybrid Search: Keywords Plus Meaning
Read: Hybrid Search: Keywords Plus Meaning
The Exact-Match Lane
Weighted Fusion
Sidebar: Lexical Scoring (BM25) in One Picture
Further Reading
Next
Doodle Cheat Sheet: Hybrid Search: Keywords Plus Meaning
Cheat Sheet at a Glance
Exercise: Tell the twins apart
Check Your Work
Quiz: Hybrid Search
99 attempts allowed
Module 8: Reranking: The Second Opinion
Watch: Reranking: The Second Opinion
Read: Reranking: The Second Opinion
The Two-Stage Funnel
The Cross-Encoder Judge
Further Reading
Next
Doodle Cheat Sheet: Reranking: The Second Opinion
Cheat Sheet at a Glance
Exercise: Find the smallest net
Check Your Work
Quiz: Reranking
99 attempts allowed
Module 9: Query Understanding: Rewrite, Route and Decompose
Watch: Query Understanding: Rewrite, Route and Decompose
Read: Query Understanding: Rewrite, Route and Decompose
The Query Rewriter
The Query Router
Sidebar: Hypothetical-Answer Search (HyDE)
Further Reading
Next
Doodle Cheat Sheet: Query Understanding: Rewrite, Route and Decompose
Cheat Sheet at a Glance
Exercise: Route a mixed-language ticket
Check Your Work
Quiz: Query Understanding
99 attempts allowed
Module 10: When Sources Disagree: Authority, Applicability and Abstain
Watch: When Sources Disagree: Authority, Applicability and Abstain
Read: When Sources Disagree: Authority, Applicability and Abstain
When Sources Disagree
Abstain Thresholds Across Sources
Sidebar: Citation Formats
Further Reading
Next
Doodle Cheat Sheet: When Sources Disagree: Authority, Applicability and Abstain
Cheat Sheet at a Glance
Exercise: Out-rank the newer ticket
Check Your Work
Quiz: When Sources Disagree
99 attempts allowed
Module 11: Context Assembly: Packing Evidence for the Model
Watch: Context Assembly: Packing Evidence for the Model
Read: Context Assembly: Packing Evidence for the Model
The Evidence Pack
When Evidence Distracts
Further Reading
Next
Doodle Cheat Sheet: Context Assembly: Packing Evidence for the Model
Cheat Sheet at a Glance
Exercise: Sweep k
Check Your Work
Quiz: Context Assembly
99 attempts allowed
Module 12: Structured Knowledge: Query Specs, Scope and Safe SQL
Watch: Structured Knowledge: Query Specs, Scope and Safe SQL
Read: Structured Knowledge: Query Specs, Scope and Safe SQL
Question to Query
Scope Before You Query
Sidebar: The Read-Only SQL Guard
Further Reading
Next
Doodle Cheat Sheet: Structured Knowledge: Query Specs, Scope and Safe SQL
Cheat Sheet at a Glance
Exercise: Find my final-sale orders
Check Your Work
Quiz: Structured Knowledge
99 attempts allowed
Module 13: Knowledge Graphs and Multi-Hop Retrieval
Watch: Knowledge Graphs and Multi-Hop Retrieval
Read: Knowledge Graphs and Multi-Hop Retrieval
Entities and Edges
The Multi-Hop Path
Further Reading
Next
Doodle Cheat Sheet: Knowledge Graphs and Multi-Hop Retrieval
Cheat Sheet at a Glance
Exercise: Three kinds of answer
Check Your Work
Quiz: Knowledge Graphs and Multi-Hop Retrieval
99 attempts allowed
Module 14: Index Lifecycle: Updates, Deletes, Rebuilds and Drift
Watch: Index Lifecycle: Updates, Deletes, Rebuilds and Drift
Read: Index Lifecycle: Updates, Deletes, Rebuilds and Drift
The Stale Index
Build, Check, Switch
Sidebar: Incremental Re-Indexing
Further Reading
Next
Doodle Cheat Sheet: Index Lifecycle: Updates, Deletes, Rebuilds and Drift
Cheat Sheet at a Glance
Exercise: Change only who may see it
Check Your Work
Quiz: Index Lifecycle
99 attempts allowed
Module 15: Evaluating RAG: Retrieval Metrics and Faithfulness Judges
Watch: Evaluating RAG: Retrieval Metrics and Faithfulness Judges
Read: Evaluating RAG: Retrieval Metrics and Faithfulness Judges
The RAG Metrics Map
The Full RAG Scorecard
The Faithfulness Judge
Sidebar: Judge Bias (Position, Length)
Further Reading
Next
Doodle Cheat Sheet: Evaluating RAG: Retrieval Metrics and Faithfulness Judges
Cheat Sheet at a Glance
Exercise: Swap the order
Check Your Work
Quiz: Evaluating RAG
99 attempts allowed
Module 16: RAG Test Sets and Regression Gates
Watch: RAG Test Sets and Regression Gates
Read: RAG Test Sets and Regression Gates
The Synthetic Question Factory
The Regression Gate
Further Reading
Next
Doodle Cheat Sheet: RAG Test Sets and Regression Gates
Cheat Sheet at a Glance
Exercise: Pass the knowledge exam
Check Your Work
Quiz: RAG Test Sets and Regression Gates
99 attempts allowed
Wrap-Up
You Built the Whole Knowledge System
Get Your Certificate
Where to Go Next
AI Engineering Visualized: RAG, The Course
Build RAG answers you can trace to a source, module by module: over 2 hours of video, 100+ doodles, 16 exercises, 16 quizzes and a certificate.
Minimum price
$34.99
$49.99
You pay
Author earns
About
About the Course
Your search found a match. It was an old ticket that said a delivered parcel was still on its way. A better search would still find that ticket: the code around the search decides which sources may answer, and that's what this course builds. You pick up the support assistant for a fictional store, Northwind Goods, and give it real knowledge: source contracts, ingestion with stable ids and provenance, documents with tables and scans, chunks that know where they apply, access checks, hybrid search, reranking, query routing, abstain floors, evidence packs with checked citations, safe SQL, a knowledge graph, index syncs, evals and a regression gate. Each module gives you a short video, a reading lesson, a one-page doodle cheat sheet, a small coding exercise and a quiz. The code runs on Python with a built-in stub model, so you don't need an API key. One knowledge system, built module by module. I wrote this course for developers who have built a retrieval feature and now have to trust its answers: from manuals with tables, from an order database, from past tickets and from rules that change by region. Retrieval only proposes evidence. The code around the search decides which sources may answer, and that's what we draw and build here, one part at a time. What you'll build You start with an assistant whose best search match is a stale ticket. By the end, your knowledge system: • gives every source a contract: which questions it may answer, its authority, freshness, owner and scope • ingests with stable ids, provenance and a manifest, so reruns are safe • keeps table rows as records and reads the page image when the layout carries the meaning • cuts chunks at headings and rows, and tags each one with where it applies • takes identity from the session and checks access at search, expansion, rerank, cache and citation • picks its embedder on your own golden set, and puts exact product codes ahead of the fused ranking • retrieves wide, reranks to five, and rewrites, routes and splits questions before search • decides which source wins when two disagree, and abstains below a calibrated floor • packs evidence the model can cite, and validates every citation • answers from records with allow-listed, parameterized SQL, and follows multi-hop paths through a knowledge graph • keeps the index in sync with updates, deletes and rebuilds, and is graded with retrieval metrics, a faithfulness judge and a regression gate What's inside • 16 modules, plus a short Start Here module that sets up the code • 16 video lessons, about 8 minutes each, so over 2 hours of video in all • 100+ hand-drawn doodles, each one a named mental model like the Knowledge Sources Map or the Regression Gate • 16 one-page doodle cheat sheets • 16 coding exercises, each with a prediction question, the expected output and a worked solution • 16 quizzes. Conceptual questions only. Retake as often as you like; your best result counts. • A certificate of completion: pass with 70% or more, or 90% or more for a distinction How each module works Watch the video. Read the lesson. Keep the cheat sheet. Do the exercise: predict what the code will print, run it, compare. Take the quiz. That's it, at your own pace, with no start date and no deadline. The code Everything runs on Python 3.10 or newer, with git and one requirements file to install. A built-in stub model means you don't need an API key, a network connection or a paid account, and every exercise has an expected output you can check against. If you'd like to try a real model, you can point a chapter at Ollama on your own machine. Who it's for Developers who can read Python and have taken Volume 1, or covered the same ground: embeddings, search, chunking, a RAG pipeline, grounded answers and evals. The code starts from Volume 2's final code; where a chapter uses Volume 2's harness, it's recapped in a line. You don't need a machine learning background. Who it's not for If you want a tour of vector databases or RAG frameworks, or want to train models, this isn't that course. It builds plain code so that every check is code you can read. Modules 0 Start Here · 1 The Knowledge Map · 2 Ingestion Pipelines · 3 Documents With Layout · 4 Structure-Aware Chunking and Applicability Metadata · 5 Knowledge Access · 6 Embedding Models and Vector Indexes · 7 Hybrid Search · 8 Reranking · 9 Query Understanding · 10 When Sources Disagree · 11 Context Assembly · 12 Structured Knowledge · 13 Knowledge Graphs and Multi-Hop Retrieval · 14 Index Lifecycle · 15 Evaluating RAG · 16 RAG Test Sets and Regression Gates The course and the book The course follows the book AI Engineering Visualized, Volume 3: RAG & Knowledge Systems (Vol 3), chapter for chapter. If you already have the book, the course adds the videos in each module, the one-page doodle cheat sheets, the prediction questions, the quizzes and the certificate, with your progress tracked as you go.
Instructor
About the Instructor
Prateek Singh is the solo technical founder of Ganak AI Labs, an Indian AI SaaS startup that also runs AI workshops for builders. Its product, Cloudeval AI, runs pre-deployment reviews for Azure, plus AWS CloudFormation (beta), with evidence-backed reports, through the web, a CLI, GitHub and MCP. Before AI, he worked as an infrastructure and cloud developer and was an active PowerShell community writer.
Books and courses: AI Engineering Visualized (Vols 1–4), Learn Windows Subsystem for Linux (Apress, 2020), The UX Foundations of Agentic AI Systems, PowerShell to C# and back, Learn C# in 30 Minutes, PowerShell Guide to Python, Plain Perception, and the Educative course Learning Python through PowerShell Concepts.
Material
Course Material
The Leanpub 60 Day 100% Happiness Guarantee
Within 60 days of purchase you can get a 100% refund on any Leanpub purchase, in two clicks.
See full terms...
Earn $8 on a $10 Purchase, and $16 on a $20 Purchase
We pay 80% royalties on purchases of $7.99 or more, and 80% royalties minus a 50 cent flat fee on purchases between $0.99 and $7.98. You earn $8 on a $10 sale, and $16 on a $20 sale. So, if we sell 5000 non-refunded copies of your book for $20, you'll earn $80,000.
(Yes, some authors have already earned much more than that on Leanpub.)
Learn more about writing on Leanpub
Free Updates. DRM Free.
If you buy a Leanpub book, you get free updates for as long as the author updates the book! Many authors use Leanpub to publish their books in-progress, while they are writing them. All readers get free updates, regardless of when they bought the book or how much they paid (including free).
Most Leanpub books are available in PDF (for computers) and EPUB (for phones, tablets and Kindle). The formats that a book includes are shown at the top right corner of this page.
Finally, Leanpub books don't have any DRM copy-protection nonsense, so you can easily read them on any supported device.
Learn more about Leanpub's ebook formats and where to read them
Write and Publish on Leanpub
You can use Leanpub to easily write, publish and sell in-progress and completed ebooks and online courses!
Leanpub is a powerful platform for serious authors, combining a simple, elegant writing and publishing workflow with a store focused on selling in-progress ebooks.
Leanpub is a magical typewriter for authors: just write in plain text, and to publish your ebook, just click a button. (Or, if you are producing your ebook your own way, you can even upload your own PDF and/or EPUB files and then publish with one click!) It really is that easy.