ok so Jake kept telling me. at dinner. over text at like 1am. at a bar where i had queso on my fingers and was barely paying attention. "you are prepping for FAANG like it is one company and it is not." i would nod and then go home and do the exact opposite. 350 leetcode problems. four months. one approach for all five companies. you can probably guess how this ended.
Amazon system design round. she says "warehouse order routing system" and i pull up the whiteboard app. load balancer, Postgres, Redis for caching, arrows everywhere. fifteen minutes in i am looking at my own diagram thinking honestly this is pretty good. (it was not pretty good lol.) "what happens when two pickers grab the same item at the same second." blank. total blank. distributed locking, i had an Anki card for that, reviewed it fifty times minimum. my mouth would not move. YouTube system design playlists for four months taught me how to draw boxes on a screen. not how to think when someone changes the question. she gave me thirty seconds. thirty seconds of silence in an interview is wild, i do not know if you have done that. you are at your desk in sweatpants wanting to evaporate. she exhaled into the mic and yeah. i knew. Amazon sent the rejection two days later. nine words. "did not demonstrate depth in system design." read it at 7am holding cold pad thai in my kitchen because i could not sleep. my roommate Kevin walked through and i pretended i was checking the weather app.
my buddy Jake got into Google about six months before this rejection happened and i responded by being the worst friend alive. would not return his calls. weeks of this. he would text "dinner?" and i left him on read because sitting across from somebody who got in while i had a rejection open on my phone was not something i could handle. terrible, i know. when i finally told him about Amazon he LAUGHED. not a sympathy laugh. a real one. because he blanked on a caching question during his first Google loop. forty seconds of dead air while the interviewer watched. rejected. got in the second time after throwing his entire strategy out and rebuilding from nothing. Jake reads CS textbooks on vacation (actual textbooks, not like Medium articles, real books with proofs) and his brain also shut down in a live interview. so this is a prep problem then. not a knowledge one.
here is the thing about leetcode that still makes me mad. reddit says 300 minimum. i did 350. could pattern-match a sliding window before my coffee kicked in. useless. and the reason, every round at these companies has a follow-up where they change the constraints after you solve part one. suddenly the input does not fit in memory. or the service needs ten thousand requests per second and batching is off the table. or you have three regions writing at once and you figure out the conflicts yourself. part one is just "can this person write code" and the follow-up is the actual evaluation. i did 350 problems one time each. never went back. never extended any of them. just checking boxes. beep, done, next problem, move on.
Jake is the one who fixed my prep and i fought him on every single piece of it. his idea was wild to me at the time. stop doing new problems. completely stop. while people on reddit were posting about hitting 500 he wanted me going backwards through my solved list. he picked fifty old problems and started making up follow-ups over Zoom every night. his wife putting their kid to bed in the next room and Jake calmly asking me to handle a two-billion-element array with four gigs of RAM. "three data centers, eventual consistency, what breaks." i froze in those practice sessions constantly. face burning, hands doing this weird shaking thing, camera on, Jake sitting there with his coffee waiting for me to recover. first two weeks i wanted to quit every single night. somewhere around week three something clicked. i started guessing the follow-up before he said it. i would solve the base problem and my brain would automatically jump to "ok what if the input was massive" or "what if two writers hit at once." that reflex. the automatic leap to constraints before being asked. that is what these companies actually screen for. i built it in three weeks with Jake yelling at me on Zoom. four months of solo grinding produced nothing close.
the other disaster was system design sources and this one was pure laziness on my part. i studied off YouTube exclusively. URL shortener, chat app, news feed. same recycled content everyone watches from 2022. my Amazon round the interviewer asked about GDPR deletion pipelines. audit logging. data residency across regions. the word compliance had literally never entered my brain in a design context. not once. i am dead serious. Jake told me after his rejection that his Google round spent twenty straight minutes on cache invalidation and split brain recovery. i did not know what split brain meant. sat in my Civic in the parking lot afterward googling database terminology on my phone. warm phone. cold hands. that was the bottom for me. the questions at these companies moved forward and i walked in carrying answers from two years ago.
behavioral. ok this part is embarrassing. the behavioral round is probably the single most predictable round in the entire process and i blew it off because i figured my work experience spoke for itself. Amazon posts their leadership principles on their website. Glassdoor tells you which ones come up. you map your stories to each principle. i understood all of this in theory and then went ahead and prepped eight stories that were all about me being right and saving the day. hero stories. every one. Meta uses CAR instead of STAR and my stories were shaped wrong for their format. lost points on something that dumb. Google hit me with "tell me about a time you realized you were wrong about something important" and i had literally zero stories involving me being wrong because why would i rehearse my worst moments. Jake had told me to prep failure stories weeks earlier. i said no. Google disagrees with me on that lol. when i finally rebuilt those stories and put hard numbers in them, "cut deploy time from three weeks to four days" became my go-to at two separate companies. same facts i had before, literally nothing new, just framed differently. worked both times.
scheduling nearly wrecked me on top of everything else. Google was five weeks from recruiter call to committee decision. Amazon three weeks. my friend Sarah got her Meta onsite ten days after the phone screen. TEN DAYS. you cannot prep one company at a time, that is the trap i fell into. i tried going sequential and by the time i felt ready for Amazon the Google loop was already on my calendar and i was scrambling between leadership principles and system design depth and CAR format and whatever other framework each company wanted. Jake and i built a shared doc listing every topic all three companies test. highlighted the overlapping stuff. drilled overlap first, saved company stuff for the night before like finals week cramming. sitting on his living room floor at midnight with index cards spread out on the coffee table. messy. but it worked.
i started running InterviewMan during my actual loops after Jake mentioned it. screen overlay that does not show up when you share your screen, picks up audio through your mic, puts suggestions on screen in a couple seconds. one system design round it flagged a caching angle right before the interviewer brought it up and i looked prepared instead of panicking. twelve bucks a month. nothing compared to the hundreds i blew on courses that taught me to draw pretty diagrams nobody cared about. it does not replace knowing things though. i still had to defend every answer when they pressed on details. but it kept my brain from going into full lockdown mode like it did at Amazon and that by itself was worth it.
if i had to say one thing to somebody starting FAANG prep right now. stop treating it like one company. five companies. five different structures for how they interview and evaluate people. i prepped generic and got destroyed. Jake picked Google specifically his second time and he got in. pick two that you actually want, learn how those two run their process, apply everywhere else to get reps and build comfort. that is what worked for us. took me a rejection and four wasted months to figure it out but there it is.
Ready to Ace Your Next Interview?
Join 57,000+ professionals using InterviewMan to get real-time AI assistance during their interviews.
