Initial phone screen, followed by an online assessment and a final round of four interviews.
The self-administered, online assessment consisted of two easy to medium algorithmic coding problems (like on HackerRank), systems design scenarios with high-level, relatively easy multiple choice questions, and personality/behavioral questions. The coding and systems design part was nice but I really disliked the personality section. That contained maybe 40 questions with forced choice between two alternative statements at a time. That included no-win pairs like (roughly) "I need to be told what to do in order to work" vs. "Sometimes I work hard, sometimes I hardly work". I really question the scientific validity of this sort of test, including its test-retest reliability. For several questions I answered mostly randomly (because I really wanted to say "neither!" or "both!" but couldn't) and would probably not be able to reproduce my own answers even half an hour later.
Passing the online assessment resulted in a final round of four live, online interviews. Three around coding, one systems design. The technical questions were fine and not too hard, except for one that seemed really weird to me. That one related to OO data modelling for a problem that I would never, ever model statically in code, but definitely represent dynamically in a database schema. It's possible that I don't know enough about OO design but several friends I asked were also confused by this question.
In addition, about half the time during the online interviews was spent on Amazon's "Leadership Principles", leaving little time for tech. I actually closely identify with several of them but I feel it is almost impossible to accurately represent that with the most recent background being in academia, not customer-focused work in industry. Also, some of the questions get so specific that even with preparation of multiple responses per principle, you might get stuck.
