AI engineer interview: the learning map
Project-based tasks across six spheres, the areas an interview loop samples. Each task is a ladder of rungs, versions of one program that go from a toy to a production shape, with worked rungs you stop needing. The goal is to do it, explain it and defend it with no assistant in the room.
The scaffold level is how much help a task page shows.
Applies to every task page. Worked code at 0; hints at 1; target and sources at 2; nothing but the questions at 3. Tests and feedback are the same at every level.
How this was designed: the needs analysis and the research behind it
Spheres
Every task is tagged with the spheres it exercises. Interview rounds map onto these; a gap in the map is a gap in the loop.
- software engineering
- programming language
- API
- AI
- mathematics and theory
- production and scale
Tasks
How a task is built, in order
- A realistic scenario with a toy version that runs in one sitting.
- Worked rungs, each measured by a contract test, a test that pins what the rung costs, so you predict before you run.
- Curated sources with what each is for; every claim on a page cites one.
- Five-minute interactive lessons for the waiting-room slots.
- A learner target that is red until you make it green, at the scaffold level you chose.
- Cited feedback on your code and your explanation (
make feedback). - Interview questions the task answers, and the one it leaves open.
New tasks are added when a need shows up, not in advance. The plan and the backlog live in the repository README.