
An ML engineering loop is not one interview repeated five times. Each round grades something different, and the mistake that costs the most offers is preparing for the coding round and treating the other four as conversation.
This round is a filter, and it is looking for two things: that your experience matches the posting closely enough to justify the team's time, and that you can describe it without rambling. Have a ninety-second version of your last two years ready. Most people have a five-minute version and get cut before the interesting part.
For ML roles this splits in two. Some companies run a standard algorithms interview. Others ask you to implement k-means, or attention, or beam search from memory. Prepare for both. The second kind is far less commonly practiced and far more revealing.
Rapid questions across the whole field with no warning about which area. What is being graded is calibration: whether you can give a two-sentence answer when that is what was asked, and go three levels deeper when the follow-up wants it. Candidates who monologue on every question score badly even when every answer is correct.
The round that decides your level. Forty-five minutes, a vague prompt, and a whiteboard. The structure matters more than any individual choice:
Running out of time before the tradeoffs is the single most common way this round is lost.
It is graded, and vague answers cost offers. Build a story bank with real numbers in it and stress-test each story against three follow-ups, because that is what a trained interviewer will do.
Audit yourself round by round rather than in aggregate. Almost everyone has one round that is quietly weak, and it is almost never the one they are practicing.

When the answer is wrong, teams reach for a better model. The document usually was never retrieved in the first place.

Loss flat, loss spiking, training down and validation up. Each shape has a short list of causes, and guessing is not one of the steps.

Reaching for a neural network on tabular data is a common way to look junior in a system design interview.