Practical AI training for everyone
Module 1
9 min read
"AI job" describes at least four different careers with different daily work, different interviews and different entry routes. People waste months preparing for the wrong one. The first useful thing this track can do is tell you which door you are standing in front of.
There is also a fast-growing fifth category with no settled name: the person inside a non-technical organisation who understands these tools well enough to lead their adoption. It pays less than the engineering roles and is far easier to enter from a standing start.
Very little of an AI engineer's week is spent thinking about models. It is spent on data plumbing, on why the retrieval step returned the wrong chunk, on evaluation, on cost, and on the same code review and deployment work as any other engineering job. Candidates who expect otherwise interview badly, because they talk about capabilities when the interviewer is asking about failure modes.
A real week, roughly
Two days extending an ingestion pipeline so a new document type is chunked sensibly. One day on an evaluation set after support flagged three wrong answers. Half a day reducing token spend on a prompt that runs on every request. The rest in review, standups and a meeting about whether a feature should exist. Model selection came up once, for twenty minutes.
Two things, in this order. Evidence that you have built something that works and that you understand where it fails; and the ability to reason out loud about a system you have not seen before. Certificates matter less than either. That is the honest reason this track ends in four projects rather than a tenth quiz.
Try it now
Find three job adverts that interest you and classify each into one of the four families above. Then list, for each, the specific requirement you cannot currently evidence. That list is your actual syllabus, and the rest of this track is designed to close most of it.
A few quick questions. You'll see an explanation after each one.
Question 1 of 5
Which family covers most newly created 'AI engineer' roles?