So testing it is a different job. Sixty-three short lectures on how these systems actually go wrong - bad data going in, behaviour that quietly drifts once it is live - and what you do about it. The ISTQB CT-AI certificate comes with the territory, if you want the paper.
Fifteen minutes a day, and the questions get mean on purpose.
One idea, five to fifteen minutes, explained the way you'd explain it to a colleague over lunch. No ML degree assumed.
Straight after, while it is still fresh. Get one wrong and it tells you why, not just "nope, try again".
Stuck at 11pm on what "fairness" even means for your model? Ask Kai in Telegram. He answers, any hour.
There is a full mock exam to drill against. You book the real ISTQB one yourself - the paper is a nice bonus on top of knowing your stuff.
Thirteen modules: an intro, eleven that follow the official ISTQB CT-AI syllabus, and a final assessment. It opens with what these systems actually are and where they go wrong, then moves into how you test for that. One lecture a day; the mock exam at the end runs under real timed conditions.
Full syllabus →
I'm a QA lead, not an AI researcher. Someone handed me "the new AI thing" and I had to work out how to test it.
Twelve years in QA, ISTQB Foundation since 2018, and these days I work alongside about twenty other testers. When AI landed on my desk I started building the course I wish someone had handed me. It took two years and around €35,000.
It is theory first: what these systems actually are, and what words like fairness and drift come to mean once the thing is learning on its own. The CT-AI certificate sits on top, for when you want it. The theory is the part that makes people in an AI standup stop and actually listen to you.
