Analyst interviews often include a short live-coding task in plain Python before any pandas. Most of these questions reuse patterns you've already practiced.
Counting and ranking means Counter, .most_common() and sorted with a key. Many questions get easier when you look things up instead of searching: build a set or a dictionary once, then check membership in one step. That's how the classic "find two values that add up to a target" question is solved.
Edge cases come up on purpose: an empty list, ties, duplicates, a single value. The hidden checks here test them too. And say what each step does as you write it, because the interviewer is listening for your reasoning as well as the result.
Each exercise is a function, checked on the inputs in the prompt and on hidden ones. Try each one without hints first, then read the explanation even when you get it right, since it explains the choice the way you'd out loud.