went through the Zoom SWE interview process earlier this year for a mid-level role (targeting L4). writing this up because the online assessment format caught me off guard and I couldn't find good info beforehand.
the OA: it's on Coderpad, not HackerRank or Codility. two problems, 90 minutes. you can use any language. I used Python.
problem 1 was a medium difficulty array/sliding window thing. pretty standard. I had it done in 20 minutes.
problem 2 was trickier. it was a graph traversal problem that looked like a tree problem at first, which I think was intentional. I almost coded a simpler solution before re-reading the constraints and catching that cycles were possible. spent about 40 minutes there and got it working but the solution wasn't as clean as I wanted.
what they're testing: correctness first, then time complexity. I had a comment in my code noting the O(V+E) complexity and the interviewer later said that stood out. they also want to see you handle edge cases without being prompted. empty input, single element, etc.
difficulty: roughly LeetCode medium. I did not see a hard-level problem. a friend who went through the L5 loop said one of her coding rounds was harder, borderline hard territory, so level probably matters.
the live coding round (onsite): separate from the OA. this one is 45 minutes, interviewer present, can ask questions. felt more like a collaborative problem-solving session than a pure test. the interviewer gave a nudge when I was going down a bad path, which I appreciated.
bottom line: if you're comfortable with mediums and can explain your reasoning out loud, the coding rounds are manageable. don't overthink it.