ran InterviewMan through the full amazon virtual loop (5 rounds) - round-by-round review
ok so i did the amazon virtual loop a couple weeks ago, sde2, and ran InterviewMan in stealth on a second device for all 5 rounds. got the offer. wanted to write this because i lurked here for like a year before i pulled the trigger and the round-by-round walkthrough was the thing i wish i had had before going in
setup first because thats the part everyone asks about. macbook on my desk doing audio capture, ipad on a stand to my left at maybe a 45 deg angle, that's the client where the answers stream. the desktop window is hidden, not in the dock, doesnt show up in cmd-tab, and most importantly does not show up if you screenshare. i tested with quicktime screen recording before going in just to make 100% sure nothing was leaking. nothing was leaking
loop itself was 5 rounds, all virtual on chime, all in one day. two coding, one system design, one behavioral pure on leadership principles, then bar raiser at the end. they gave you like 10 mins between each round which was barely enough to pee and stand up
starting with the behavioral / LP one because thats where this thing more than paid for itself imo. amazon hits you with maybe 6-8 LP prompts back to back and you're supposed to STAR every single one without sounding like you're reading from a script which is, you know, sort of impossible. i had it set to STAR + medium length. it would have bullet beats up while the interviewer was still finishing the question. i kept eye contact and used the ipad as a glance-down anchor
system design was a notification fan-out service. honestly i was nervous about latency going in like would the suggestions even keep up with whiteboard chatter, but they did. i didnt read it off word for word. mostly used it to remember the second tradeoff i wanted to bring up after id covered the obvious consistency vs availability stuff
first coding was a medium graph problem. honestly i mostly solved that one on my own. the panel did catch a base case i missed before i ran my tests though, saved me like 5 min of dumb debugging probably. second coding was sliding window with a twist and that was the one i leaned on harder. streaming answer let me reason out loud while glancing at it. the interviewer asked good followups so i had to actually get the algo not just read it, which is kind of how youre supposed to be using this thing anyway imo
bar raiser was half behavioral half "tell me about a time you disagreed with a senior engineer." LP rotation again, same setup same flow. she was actually nice ngl. not the gotcha-fest people on this sub make them out to be
what worked for me: ipad off-axis so my eyes didnt drift on camera (this is huge, that was a tip i picked up from this sub actually). using STAR for behavioral helped so i was hitting beats not reading whole sentences. question detection on high. low missed nothing for me either tbh but high would throw extras and i just ignored em
what id change: dont read verbatim. like ever. if you read the suggestions out loud you sound like a robot and the bar raiser WILL clock it in 30 secs flat. thats how most people get caught imo. treat them as anchors not a script. you should be ahead of it when youre talking, not catching up
also eat lunch between rounds. i didnt. i was fading hard by the bar raiser, obvious in hindsight lol
disclosure - paid for the annual myself, no affiliation, just posting because the LP gauntlet is brutal and idk this took a real chunk of the edge off for me
anyone else done a loop recently? curious if the LP weight felt the same to you or if its shifted. my recruiter said LPs were like 60% of the call but it felt closer to 40 honestly