I am new to AI infrastructure
You write software and want the whole picture before you specialise.
Where to start
8 learning materials, 61 modules, one tree. Find your branch, open its Module 1 and read down — 1.a, 1.b, then Module 2.
How it is structured
Every learning material has the same shape, so once you can read one you can read them all.
How to study a module
- Open the module overviewEach module starts with a page that says what it covers and lists its lessons in order.
- Read the lessons in order1.a, then 1.b. Every lesson assumes only what came before it.
- Run the codePrograms are complete Go files that build and run. Change a number and run them again.
- Do the exerciseLessons end with a hands-on exercise. That is where the understanding is made.
- Build the projectWhere a material has projects, each one proves a group of modules end to end.
The tree
Grouped from the bottom of the stack to the top. Open a material to see its modules.
- Pidoku8 learning materials
- Layer 1The machineWhat everything runs on — the language and the chip.
- Layer 2Running modelsHow a model becomes a fast, cheap, observable service.
- 1Fundamentals11 lessons
- 2Computer Systems Foundations13 lessons
- 3Machine Learning Fundamentals12 lessons
- 4Neural Network Inference12 lessons
- 5LLM Inference15 lessons
- 6GPU Computing12 lessons
- 7Inference Optimization14 lessons
- 8Inference Serving Systems14 lessons
- 9Distributed Inference12 lessons
- 10Memory and Performance Engineering11 lessons
- 11Production Inference Engineering11 lessons
- 12Inference Platform Engineering11 lessons
- 13Advanced LLM Inference12 lessons
- 14Research and Frontier11 lessons
- Layer 3Designing systemsHow whole AI systems are designed, secured and sold.
Pick a path by goal
Not reading everything? These orders get you useful fastest. Each step opens at the first module of its range.
I want to run and tune LLM serving
Inference, ML platform or SRE engineers who own the latency and the GPU bill.
I design AI products and platforms
Solution architects, tech leads and anyone who has to justify a design and its cost.
I secure AI systems
Security engineers and reviewers who have to sign off on an AI feature.
I build backends in Go
Go engineers who want the runtime underneath and AI systems on top.
I care about the hardware
Anyone choosing, sizing or paying for accelerators.