System Design for AI

System design for AI, taught like a story — free, from zero to production.

A free, open-source, beginner-to-expert curriculum covering how AI systems are actually designed and built — ML, GenAI, and Agentic AI system design, taught through narrative and analogy. Plus a working Next.js website with an interactive system visualizer.

Free. Open source. MIT-licensed. No paywall, ever.

What's in this repo

system-design-for-ai/
├── chapters/              # 38 markdown chapters with YAML frontmatter
│   ├── 00-start-here/     # Module 0 (4 chapters)
│   ├── 01-ml-system-design/       # Track A (8 chapters)
│   ├── 02-genai-system-design/    # Track B (9 chapters)
│   ├── 03-agentic-system-design/  # Track C (7 chapters)
│   ├── 04-cross-cutting/          # Cross-cutting (6 chapters)
│   └── 05-design-studio/          # Capstone (4 chapters)
├── website/               # Next.js website (runnable)
│   ├── src/app/           # Pages: home, curriculum, reader, visualizer, questions
│   ├── package.json       # Dependencies
│   └── README.md          # How to run
├── 06-question-bank/      # Practice scenarios
├── assets/diagrams/       # Mermaid diagram sources
├── docs/                  # Citations, pedagogy contract
├── GLOSSARY.md            # Plain-English glossary
├── ROADMAP.md             # What's built, what's next
├── CONTRIBUTING.md        # How to contribute
├── CODE_OF_CONDUCT.md     # Community standards
├── SECURITY.md            # Security policy
└── LICENSE                # MIT

The curriculum map

38 chapters across 6 tracks. Every chapter follows Bloom's Taxonomy (Remember → Understand → Apply → Analyze → Evaluate → Create), opens with a story, and ends with an "explain it back" checkpoint.

Track	Chapters	Example
Module 0 — Start Here	4	0.0 The Friday Night Problem (What is system design)
Track A — Classical ML	8	A.2 The Front Page Mind Reader (Recommendation systems)
Track B — GenAI / LLM	9	B.2 The Librarian Who Never Forgets (RAG)
Track C — Agentic AI	7	C.1 The Loop That Won't Stop (Agent loop)
Cross-Cutting	6	X.4 The Bill Nobody Warned You About (Cost/FinOps)
Capstone Studio	4	S.1 Design a Customer Support AI

How to read the curriculum

Option 1: Read on GitHub. Every chapter in chapters/ is standalone markdown with YAML frontmatter. Diagrams are Mermaid — GitHub renders them natively.

Option 2: Run the website. The website/ directory contains a Next.js app that renders all chapters with an interactive system visualizer, searchable curriculum map, and question bank.

cd website
npm install
npm run dev
# Open http://localhost:3000

The website

The included Next.js website provides:

Landing page — overview, stats, track summaries
Curriculum map — all 38 chapters, searchable and filterable by track
Lesson reader — renders chapter markdown with Mermaid diagram support, sidebar navigation
System Visualizer — interactive RAG pipeline walkthrough (6 steps, click through the request flow)
Question bank — 8+ scenario-framed practice questions with worked answers

Built with Next.js 16, TypeScript, Tailwind CSS, shadcn/ui.

Pedagogy contract

Every chapter obeys:

Bloom's Taxonomy in order — Remember → Understand → Apply → Analyze → Evaluate → Create
Open with a story, not a definition
One idea per paragraph, one analogy per concept, plain words over jargon
No lecture voice — coffee with a sharp friend, not a slide deck
Never assume unstated prerequisites — link to earlier chapters inline
End with "explain it back" — teaching-back is the test of understanding
Cite real sources — arXiv papers, engineering blogs, official docs
Date-stamp time-sensitive claims — pricing, model names, protocol versions

Acknowledgments

This project complements (does not duplicate):

ai-system-design-guide — the best free staff-level reference
ByteByteGo — excellent paid course (we're the free alternative)
Awesome-Agentic-System-Design — curated link list

Every explanation, analogy, and diagram is written fresh.

License

MIT. See LICENSE. Free, open, forever.

Developer

Built by Adil Shamim.

System-Design-For-AI