11 hands-on labs
Every command and expected output captured on live Azure. A reference solution for the end of every lab, and one command to catch up if you skip ahead.
RAG on Azure · 11 labs · verified on live Azure, Oct 2026
Build a production RAG system into a real app, lab by lab
You take Fieldnotes, a small working team app, and build a complete RAG system into it on Azure: messy documents in, cited answers out, then measured, monitored, and hardened. Text and code, no video. Every command was run on live Azure before it shipped.
One-time purchase · 12 months of updates · Azure usage about $4–14 in total, on your own subscription
python3 scripts/ask.py "When do the old office badges stop working, and which vendor is doing the replacement?" ANSWER: Old badges stop working on April 14, 2026, and the replacement vendor is NorthPeak Cloud field services.[1] CITATIONS: [1] scanned-facilities-memo.pdf [2] policy-badge-access.html [3] vendors.csv (2.33s)
python3 scripts/ask.py "What is our policy on quantum computing purchases?" ANSWER: I can't find that in your documents.
pytest -m eval tests/eval/test_eval_gate.py ... [100%] 3 passed, 38 deselected in 190.78s
Built around the failures demos skip
Most tutorials stop when the demo answers one question. Real systems fail in ways those demos never show. Every one of these happened while the labs were built and verified on live Azure, and each one is a step you work through.
- Documents that won't parse: scanned PDFs, broken files, tables
- Search that misses because the user's words differ from the document's
- Answers that sound right and cite the wrong page
- Quality that quietly drops after someone "improves" a prompt
- Costs nobody watches until the invoice
- A document someone planted to hijack the bot
What you build
One app, end to end, the way a team would build it
- Start100 messy documentsMarkdown, PDF, Word, HTML, CSV, a scanned memo, and one corrupt file
- Lab 4Ingestion with OCRSix formats, OCR only where needed, failures visible instead of silent
- Lab 5Chunks and embeddingsSearch by meaning, batched and rate-limit safe
- Lab 6Hybrid searchAzure AI Search: keyword and vector, merged into one ranking
- Lab 7"Ask your docs"Cited, streamed answers, and an honest refusal
- Labs 8–11Run it like productionMeasured quality, a CI evaluation gate, traces and cost per answer, injection defenses
Azure OpenAI · Document Intelligence · AI Search · Application Insights. All keyless: your app signs in with your Entra ID identity, with no API keys anywhere.
The labs
11 labs, about 22–32 h
| # | Lab | What you build | Time | Azure cost |
|---|---|---|---|---|
| 01 | Run the app and meet the corpus | the stack running on your machine; a tour of the code and the 100 documents | 60–90 min | $0 |
| 02 | Your first LLM call | a keyless Azure OpenAI client; the Summarize feature | 2–3 h | ≈ $0.50–2 |
| 03 | Prompting that works in apps | streaming answers; structured-output auto-tagging | 2–3 h | ≈ $0.50–2 |
| 04 | Document ingestion: the real pipeline | six-format ingestion, OCR for scans, visible failure | 2.5–3.5 h | ≈ $0 |
| 05 | Embeddings: search by meaning | chunking, batched embeddings, semantic search | 2–3 h | ≈ $0.10–0.50 |
| 06 | Azure AI Search: the production store | the search index; hybrid keyword + vector search | 2–3 h | ≈ $0–1 |
| 07 | RAG v1: "Ask your docs" | cited, streamed answers with a testable refusal | 2–3 h | ≈ $0.50–2 |
| 08 | Retrieval quality: measure, don't argue | a quality harness; ranker and chunk-size experiments | 2–3 h | ≈ $0.50–2 |
| 09 | Evaluate the RAG like production | an evaluation gate for CI, with a model as judge | 2–3 h | ≈ $0.50–2 |
| 10 | Observability | traces, and the token and dollar cost of every answer | 2–3 h | ≈ $0.50–1 |
| 11 | Hardening & responsible AI | injection defense in layers; token budget; cache | 2.5–3 h | ≈ $0.50–1 |
| Total | ≈ 22–32 h | ≈ $4–14 |
Azure costs are estimates. Actual costs vary with how long you keep resources running, your region, and Azure's current prices. Every lab ends with a teardown that brings your run-rate back to $0/day.
What's included
More than the labs
The AI Engineer's Handbook
65 topics, from tokens and context windows to agents and Azure's AI services. Each is tagged beginner, intermediate, or advanced, with links for further reading.
A capstone with an autograder
One more feature, built to a written contract with no steps given. A grader checks your work offline. No solution ships, so the design is yours to defend.
An interview pack
62 questions interviewers ask about LLMs and RAG, with strong answers tied to what you built.
A troubleshooting index
Every error these labs can produce, on one page, with the cause and the fix.
Flashcards and updates
The 52 key terms as a deck for Anki. And 12 months of updates: fixes and re-verified editions, listed in the public changelog.
Who it's for
Who it's for
- Developers (backend, devops, data) who want real LLM and RAG skills, built the way a production team builds them
- ML and data people who've done notebooks but not production retrieval, evaluation, and operations
- Anyone preparing for AI-engineering interviews. You'll have built every piece yourself, and each lab ends with the questions you should be able to answer
Who it's not for
- Complete beginners to programming. You should be comfortable with Python and a terminal
- People who want video lectures. It's text and code: searchable, offline, at your own pace
Why this, not a video course or a chatbot
Verified, measured, maintained
Verified, not assumed
Every command was run on live Azure, and every expected output in the labs is a real capture.
One real system
You build one app end to end, the way a team would, instead of collecting disconnected snippets.
Measured
Retrieval metrics, an evaluation gate, and the cost of every answer: numbers you can quote.
Maintained
Dated editions and a public changelog. When Azure changes, the labs change with it.
2027 Edition · early access
RAG on Azure
$149 $99 USD, one-time
Early-access price. It goes to $149 at full release.
- 11 labs, with reference solutions for every lab
- The AI Engineer's Handbook, 65 topics
- Capstone with autograder
- Interview pack, troubleshooting index, flashcards
- 12 months of updates; permanent access to this edition
- Delivered as a private GitHub repo invite within minutes
All sales are final: your GitHub invite goes out the moment you pay. Try the free lab first; it's written exactly like the paid labs. If the invite never arrives, or you're charged twice, we refund in full.
Early access: all 11 labs were verified on live Azure in October 2026. Fixes ship through the changelog.
Questions
What will Azure cost me?
About $4–14 in total, if you tear down when each lab says so. Most services bill per use and cost nothing while idle. The one that bills by the hour is the paid search tier in lab 8, about $0.11 an hour, and every lab ends with a teardown that brings you back to $0/day. These are estimates; your region and Azure's prices change them.
What do I need to know first?
Python basics and comfort in a terminal. No prior LLM, embeddings, or Azure AI experience. You need an Azure subscription with billing enabled; lab 1 runs entirely on your machine.
Is there video?
No, by design. Text and code is searchable, works offline, and every command is exact.
Will this get me a job?
No course can promise that, and we don't: no promise of employment, interviews, certification, or income. What you get is the skill, and a working system you built yourself and can explain in detail. See the terms.
What about AWS or Google Cloud?
This package is Azure-only, by design. The ideas carry over; the commands don't.
How do updates work?
Your purchase includes 12 months of updates: fixes, re-verified commands, and new material, all listed in the changelog. You keep permanent access to the edition you bought.
How is it delivered?
After checkout you get an invite to a private GitHub repository within minutes. You'll need a free GitHub account.
Can my company pay?
Yes. Checkout runs through Polar, the merchant of record, so you get a proper invoice for expenses. For a team purchase, email support@labwrights.io.
What if Azure retires a model the labs use?
Lab 2 shows how to check retirement dates and switch models with one setting. Updates re-verify the labs when Azure changes.