A structured, chapter-by-chapter Claude skill that teaches AI Engineering by Chip Huyen (O'Reilly, 2024) — one bite-sized lesson at a time.
AI Engineering Tutor is a custom Claude skill that turns Chip Huyen's landmark book into an interactive, Socratic learning experience. Instead of passively reading, you get:
- 📖 ~40 bite-sized lessons (~10 min each) spanning all 10 chapters
- 🎨 Inline diagrams drawn fresh for every concept — no need to flip through the book
- 🃏 Flashcards (interactive + Anki-importable TSV)
- 📝 Lesson notes saved as Markdown files
- 🧪 Interactive quizzes with immediate feedback and explanations
- 💬 Socratic check-ins — the tutor asks, you answer, it corrects and moves on
The skill is built around the book, not as a copy of it. Every explanation is in the tutor's own words, with fresh examples and analogies — you bring the book for the prose; the tutor builds your understanding.
git clone https://github.com/YOUR_USERNAME/ai-engineering-tutor.git
cd ai-engineering-tutorClaude Projects let you attach files that Claude reads at the start of every conversation — perfect for a persistent course you can resume anytime.
- Go to claude.ai and click New Project
- Open Project Settings → Add content
- Upload these three files from the repo:
SKILL.md references/curriculum.md references/figures.md - Start a conversation inside the project and type:
/ai-engineering-tutor
That's it. Claude will read the skill files and begin the course.
Why Projects? Your conversation history is preserved across sessions, so the tutor can find your last lesson and resume. It's the closest experience to a real course.
The fastest way to activate this skill in any Claude conversation is with its slash command:
/ai-engineering-tutor
Type this at the start of any message and Claude will load the tutor immediately. You can combine it with your first instruction in one go:
/ai-engineering-tutor teach me from the beginning
/ai-engineering-tutor start at lesson 6.1
/ai-engineering-tutor quiz me on RAG
/ai-engineering-tutor where did we leave off?
Once the skill is active in a session, you don't need to type
/ai-engineering-tutoragain — just usenext, lesson IDs, or any command directly.
Once the skill is loaded, use any of these prompts to begin:
| What you want | What to type |
|---|---|
| Invoke the skill | /ai-engineering-tutor |
| Start from the beginning | Teach me AI engineering or Start at lesson 1.1 |
| Jump to a specific lesson | next 4.2 — goes directly to lesson 4.2 |
| Jump to a topic | Give me a lesson on RAG or Teach me about finetuning |
| Resume where you left off | Where did we leave off? or Continue from lesson 3.2 |
| Get a recommended path | I'm a backend dev building a chatbot — where should I start? |
next is your primary navigation command — it does two things depending on how you use it:
| What you type | What happens |
|---|---|
next |
Automatically fetches and teaches the next lesson in sequence — no need to know the ID |
next 4.2 |
Jumps directly to lesson 4.2 (or any lesson ID you specify) |
This means you never have to look up lesson IDs to move forward. Just type next after finishing a lesson and the tutor picks up exactly where the curriculum left off.
Examples:
next → loads the lesson immediately after your current one
next 4.2 → jumps to lesson 4.2 (Model Selection)
next 6.1 → jumps to the start of the RAG chapter
next 10.3 → jumps to the final lesson
Every lesson follows this rhythm:
1. Locate — lesson ID, title, and the one question this lesson answers
2. Teach — 2–4 key concepts with analogies, examples, and inline diagrams
3. Connect — why an AI engineer actually cares about this
4. Check — 1–2 questions; tutor stops and waits for your answer ← important
5. Next steps — say "next" to continue, or request notes / flashcards / quiz
The tutor stops and waits after asking questions — it won't steamroll into the next topic. Answer in your own words; the tutor will affirm, correct, and fill gaps before moving on.
Request these at any time — mid-lesson, end of lesson, or end of chapter:
make me flashcards for lesson 2.4
quiz me on chapter 5
save notes for this lesson
quiz me on RAG
make me flashcards for chapter 7
| Material | What you get |
|---|---|
| Flashcards | Interactive flippable deck inline + downloadable .tsv for Anki |
| Notes | Markdown summary file: key question, core concepts, analogy, "why it matters" |
| Quiz | 5–8 questions (multiple choice + short answer) with immediate feedback per question |
Every lesson ends with your current lesson ID (e.g. 6.4). Write it down.
When you return:
Continue from lesson 6.4
If you're using Claude Projects, the tutor can also search your conversation history automatically:
Where did we leave off?
next → automatically fetch and teach the next lesson in sequence
next 4.2 → jump directly to lesson 4.2 (any ID works)
next 6.1 → jump to the start of the RAG chapter
quiz me on [topic or chapter] → generate an interactive quiz
make me flashcards for [lesson] → generate flashcards
save notes for [lesson/chapter] → save a Markdown summary
give me a lesson on [topic] → jump to a topic by name instead of ID
what's in chapter 6? → overview of a chapter before diving in
The curriculum covers all 10 chapters of the book, broken into ~40 lessons:
| Chapter | Title | Lessons |
|---|---|---|
| 1 | Introduction to Building AI Applications | 1.1 – 1.4 |
| 2 | Understanding Foundation Models | 2.1 – 2.5 |
| 3 | Evaluation Methodology | 3.1 – 3.5 |
| 4 | Evaluate AI Systems | 4.1 – 4.3 |
| 5 | Prompt Engineering | 5.1 – 5.4 |
| 6 | RAG and Agents | 6.1 – 6.6 |
| 7 | Finetuning | 7.1 – 7.5 |
| 8 | Dataset Engineering | 8.1 – 8.3 |
| 9 | Inference Optimization | 9.1 – 9.3 |
| 10 | AI Engineering Architecture & User Feedback | 10.1 – 10.3 |
Each lesson has a unique ID (chapter.lesson, e.g. 6.4) for resuming across sessions.
- Foundation models — self-supervision, scale, multimodality, post-training (SFT, RLHF)
- Sampling — temperature, top-k, top-p, greedy decoding, test-time compute
- Evaluation — perplexity, exact match, semantic similarity, AI-as-a-judge, Elo rankings
- Prompt engineering — in-context learning, chain-of-thought, prompt injection, jailbreaking
- RAG — retrieval pipelines, BM25 vs semantic search, chunking, reranking, hybrid search
- Agents — tool use, planning, reflection, memory, compounding failure modes
- Finetuning — LoRA, quantization, memory bottlenecks, when NOT to finetune
- Dataset engineering — curation, synthesis, distillation, deduplication
- Inference optimization — latency vs throughput, KV-cache, continuous batching, parallelism
- Production architecture — guardrails, routers, caching, monitoring, feedback flywheels
- Access to Claude (claude.ai or Anthropic API)
- A copy of AI Engineering by Chip Huyen (O'Reilly, 2024) — recommended but not required to start
AI Engineering: Building Applications with Foundation Models Chip Huyen · O'Reilly Media · 2024 O'Reilly page · Chip Huyen's site
This skill teaches the book's concepts in original language and examples. It is not affiliated with or endorsed by the author or publisher. Buy the book — it's worth it.
MIT — with a note that the underlying book is copyright Chip Huyen / O'Reilly Media.