so i finally finished this process and can report back. did the coding portion twice: the initial OA and then live coding in two separate rounds during the virtual onsite. here is literally everything i remember.
Online assessment (HackerRank, 90 minutes, 2 problems):
problem 1: string manipulation. something like counting valid substrings with certain character constraints. medium difficulty, maybe medium-easy. i finished it in about 25 minutes.
problem 2: this one was harder. a graph problem, BFS/DFS, finding shortest paths with weighted edges. i got a working solution but not optimal. the test cases were strict enough that partial credit mattered.
if you're aiming to pass the OA: medium LeetCode comfort is the bar. solve 50-60 mediums spread across arrays, strings, graphs, and dynamic programming and you'll be fine.
Live coding (CoderPad, during phone screen and onsite):
phone screen was one problem, standard medium, two-pointer approach, the interviewer was helpful when i got stuck. he asked me to walk through my thinking before coding, which made it feel less like a graded test and more like pair programming.
onsite coding rounds: two separate 45-minute sessions. problems were similar difficulty to the OA. one had a follow-up where he asked me to optimize my first solution from O(n^2) to O(n log n). that part got a little tense. the other round had a debugging exercise where they gave me broken code and asked me to find the issues. that was actually kind of fun, different from pure leetcode grinding.
what i noticed: they do not use the word 'optimize' in the same way FAANG does. they care that your solution works and that you can explain it. the debugging round especially felt like it was testing real-job skills more than pure algo knowledge.
overall: if you've been grinding for big tech and striking out, ELC could be a solid place to actually land. the bar is real but the process felt human.