I think many of us would agree that job interviews (in tech at least) are horribly broken, because they don't do a good job of testing candidates' ability to do the actual work they'll be doing day-to-day. So saying exams are like job interviews is not a positive for exams. And, for most people, the ideal is to find a job and stick with it for years, so it's not like job interviews are common, everyday occurrences.
For presentations, usually you spend a lot of time preparing for them (similar to exams), building a slide deck or pages of notes that you refer to while giving the talk (not similar to exams). Sure, you do have to be able to think on your feet, but I don't think the comparison to a sit-down exam is all that apt.
For mundane work tasks, you have the internet and whatever reference materials you want (including LLMs, these days); this sort of thing is so different from a sit-down exam that it's almost comical that you'd try to equate the two.
I'm not saying I know of a better way to evaluate learning than proctored, in-person exams, but suggesting that sort of situation is particularly relevant to real life... no, no way.
Having been both a data analyst and software engineer I agree. The data analyst one? Here is 50K of Excel rows with all kinds of weirdness in it, you're data analyst right? You have 4 hours to analyze this data. Go!
The software engineer one: here is a takehome assignment. One week later: finished!
To be fair, they both represented pretty well what work I'm going to do. The data analyst didn't show that well how much I'd also be data engineering, but whatever, I was a SWE before having a DA stint. Back to SWE again though.