Learning paths for four roles
The tracks, put in order for a senior product manager, an AI product lead, an AI engineer, or an AI engineering lead. Each path says which lessons to read, what to skip, and when to move on.
You do not need every lesson. Read the path for your role, in order, and skip the rest.
Last reviewed: 2026-10 · Volatility: medium
This library has more than twenty tracks. You do not need all of them. A learning path picks the tracks and the lessons for one role, puts them in order, and says when to move on.
A note on scope. The paths do not teach anything new. They order and explain the lessons that already exist, and they say which lessons to skip. The lessons carry the content. The machine learning, tensors, CNNs, transformers, fine-tuning and AI-assisted development modules are published.
Pick your role
| Path | Who it is for | Outcome | Length |
|---|---|---|---|
| Senior PM | A product manager with several years of experience who is starting to own AI features | You can lead an AI feature end to end. You choose the approach, set the quality bar, price it, and judge the risk. | about 20.5 hours |
| AI Product Lead | A product leader who owns an AI product line or a team of PMs shipping AI features | You set direction for how the product grounds its answers, remembers, acts, stays safe and stays affordable. You can review a design with engineering and ask informed questions. | about 31.5 hours |
| AI Engineer | A software engineer who builds AI-powered systems | You can build, test and run a retrieval-backed, tool-using agent, and explain how it fails. | about 84 hours |
| AI Engineering Lead | An engineering manager or tech lead who runs a team that builds AI systems | You review AI architecture, set quality and safety standards, control cost, and work well with product. | about 42.5 hours |
Where the depth lives
Each path page lists its lessons by stage. The tracks hold the depth. This table shows which path reads how much of each track. The recap has the full matrix.
| If you want the full depth on | Read |
|---|---|
| Models, tokens and context windows | LLMs |
| Grounding a model in your data | RAG & vector databases |
| Building an agent by hand | Harness engineering |
| Quality, safety and cost | Evaluation & observability, AI security & guardrails, Cost optimization |
| The product craft | Technical product management |
How a path works
- Stages. Each path has four or five stages. A stage has a short reason, the lessons to read, and a checkpoint.
- Checkpoints. A checkpoint is something you can do or explain. If you cannot, read the stage again.
- Skip-ahead rules. Each page says what to skip if you already know a topic.
- Time. Estimates are rough. A reading lesson is about 25 minutes. A harness lesson that you build is about an hour.
- Not sure where to start? For the harness track, the
/find-your-levelskill runs a short placement quiz. For the rest, start at stage 1 of your path and use the skip-ahead rules.
If you wear two hats
Read the path for the job you are in now. Add stages from the other path when a task needs them. A product lead working with an agent team can add the agent stage of the engineer path. An engineering lead who writes code can add the build stages.
📌 Close out: Recap: paths compared, which ends with a self-test.