Open curriculum
An AI engineering curriculum, published module by module before the first lesson
63 designed modules, organised into 5 tracks and 4 levels. The programme is defined and published in full; recording begins with Phase 1. Nothing here is available to watch yet — and every module states exactly what stage it is at.
0 of 63 modules recorded. Curriculum revised on 2026-08-05.
The whole curriculum, by level
- Syllabus written
- Title only, so far
- modules in the curriculum
- 63
- with syllabus and exercise
- 18
- tracks × levels
- 5 × 4
- modules recorded
- 0 of 63
Four levels
Each level states who it is for and how many hours it is expected to take.
- Beginner100
Who it is for
Has never programmed, or programs very little; wants to understand AI
Suggested workload
24–36h
- Intermediate200
Who it is for
Knows basic Python; wants to build models and apps
Suggested workload
36–48h
- Advanced300
Who it is for
Already trains models; wants depth and production readiness
Suggested workload
32–56h
- Expert400
Who it is for
Professional; wants state of the art, research and scale
Suggested workload
38–46h
Five tracks
A track is a curated selection of shared-trunk modules, plus modules from its own area, plus one exclusive integrative project. It is the complete programme of an area, across all four levels.
- AGAgentic AILeave able to build autonomous agents that use tools, plan, and carry out real tasks.12 own modules · 14 reused
- DLDeep LearningTruly master neural networks — from the mathematics to training large models.9 own modules · 17 reused
- CVComputer VisionTeach machines to see — from image classification to real-time systems.8 own modules · 20 reused
- NLPNLP & LLMsMaster natural language, from embeddings to LLM fine-tuning and RAG.8 own modules · 17 reused
- MLOPSAI Engineering & MLOpsBecome the person who puts AI into production — the best-paid skill on the market.8 own modules · 12 reused
If forty hours already solve your problem, there is a shorter slice of the same curriculum: the ten 40-hour programmes
What already exists · What does not yet
The two halves of the diagram above, written out. Every line is a count over the curriculum registry, taken when this page was built — no figure in this section is written by hand.
What already exists
The lit nodes of the diagram.
- 63 modules with a code, an area, a level and a workload
- 18 of them with the syllabus and the exercise written
- 5 tracks × 4 levels, published in 3 languages
- 432 hours of designed workload, summed module by module
What does not yet
The hollow nodes of the diagram.
- 0 of 63 modules recorded
- 45 modules still without a syllabus — title only
- No cohort, no start date, no enrolment
- No student rating, because there has been no student
How each module works
A standard 2-to-4-hour format, roughly 30% theory and 70% hands-on work.
Lean theory
The concept, the visual intuition, the why.
10–20 min of video
Code-along
You build alongside the student, making mistakes on purpose and fixing them.
30–60 min
Hands-on challenge
Same problem, different dataset and context. No copy-paste.
1–2h of student work
Walkthrough of the solution
The official solution plus the most common mistakes.
10–15 min
Knowledge check
A short quiz plus submission of the code on GitHub.
At the end of the module
Why the curriculum is modular
Each module is a single object; tracks and programmes are collections that reference it. Update a module and it updates across all of them at once. That is why 63 modules cover five complete tracks instead of requiring a separate recording per track.
Courses and prices
The whole catalogue, in three currencies, with the real state of every module next to the value. No lesson has been recorded yet, which is why every price is marked as an estimate.
Browse the courses