just finished the Scale AI online assessment and technical phone screen this week. going to dump everything here because i spent hours looking for this info and couldn't find much.
online assessment (if you get one): hackerrank, 2 problems, 90 minutes. timed strictly. the problems i got were both medium level on the leetcode scale. one was arrays/sliding window and one involved binary search. nothing about data labeling or ai specifically, just pure algorithmic coding. python was fine, no issues.
technical phone screen (60 min): coderpad. one problem, medium-ish difficulty. mine was graph-related but the interviewer said they rotate problems. they start by asking you to talk through your approach before coding, then you implement. they'll ask you to test it with a few cases. they did ask about complexity but not in a 'gotcha' way.
things i noticed: the interviewer was calm and genuinely helpful when i got stuck. they gave me a nudge rather than just watching me spin they seemed to care that i could reason out loud, not just that i got to the right answer instantly no system design in this stage at all, that comes later in the loop
difficulty vs other companies i've done: comparable to meta/amazon phone screens. harder than stripe's initial phone screen, easier than google's. i'd say prepare medium leetcode consistently and you're probably fine for the initial rounds.
still waiting on onsite scheduling. nervous but the process felt fair so far.