What the course is for, how the exam works, and the pipeline that the rest of the material fills in: problem, model, architecture, solution.
How to derive a model from a domain, which archetypes the concepts become, and how a model keeps adhering to a domain that will not stop moving.
Meta-models, domain-specific languages and the machinery that turns a formal model into running code — with Xtext and the Sheduler language as the worked example.
The same engineering discipline applied to workflows whose artefacts are trained models and prompts: tracking, projects, registries, containers, and evaluation as the new unit test.
A real project brief, in the language of its client, read through every lens of the course: glossary, building blocks, an external DSL, storage behind an interface, and the exam requirements it does and does not already satisfy.