|
For lack of a better venue, I figured I'd bitch here about a bad Amazon interview experience. There was a systems engineering role I applied for and subsequently discussed with a recruiter. I had done similar integration kind of stuff with software automation before, so it seemed like a close fit. The recruiter said the manager wanted to see some code. I don't really have any major public code, so I doodled up a little project and threw it up on GitHub. It took me three weeks to scrape together the time to develop something on a topic I was interested in. It had to do with synchronizing independent agents on different systems using RabbitMQ. I thought it appropriate. After I sent the code, they quickly set me up with a phone interview. I assumed they saw the code and were keen to talk to me. As it turned out, they never looked at the code, and didn't have any intention of looking at the code. They just poked me about general OS administration questions like what Linux shell commands to use to determine disk throughput is getting maximized, migrating virtualization nodes, SQL commands, some details about the top command. There wasn't a drat programming question in the whole lot. Okay, I was asked what "use strict" meant in Perl. I had been working mostly in Windows recently; my job at best really only does Android, which is a real bastardization, and I only toy with it at home mostly these days. I managed to fix an Ubuntu rig at work that botched its upgrade, but it's not stuff I do every day, and I usually have the documentation right there. I said I had developed in a Linux environment before, but I wasn't an administrator by any means. So they ultimately rejected me after that. I wouldn't really care about that if it weren't for the fact that they never had any interest in the actual code. I actually am finding it really offensive. I'd say it's a good save: If they're willing to throw away projects for people before they've even hired them, I think I'd start the job already downhill. That's not cool. So I'd just just double-triple check with people asking for code what their intentions are with it and the process they are following. It might even be worth getting a link from them to something they consider "sufficient." I suspect these people would have been fine with a self-contained Perl script. I guess at the least that I can point to the GitHub permanently on my resume.
|
#
¿
Aug 3, 2016 07:24
|
|
|
|
| # ¿ Apr 26, 2024 23:21 |
|
|
Good points, but that assumes the interviewer will ever even look at the code. It was ultimately a pointless exercise. They had no intention to do it, even though they asked for it.rt4 posted:But yeah gently caress Amazon. That place is a factory for horror stories. Yeah, I'm staying away now. We do a lot of shopping on there, but now I'm terrified some technical debt is going to avalanche on top of everything.
|
|
#
¿
Aug 3, 2016 15:12
|
|
|
baquerd posted:Yeah, but the guy comes off as a pathological liar, which is a bit much ever for a software engineer. Yeah. You literally cannot trust what he gives for status and basically need a whole CI flow just to know what is really happening with him alone. Then you probably need somebody else writing all the tests--basically constantly auditing him. Sure, you can argue something close is ideal or whatever, but forming it on a foundation of distrust is not.
|
|
#
¿
Aug 10, 2016 15:10
|
|
|
Their timeline seems typical for huge corporations. I think that's broken my brain because people are complaining about their turnaround and I'm thinking, "Situation normal." I mean, it's still poo poo, but I don't know why people are calling them out on it in particular. I guess people still treat them as small and lean.
|
|
#
¿
Aug 22, 2016 18:06
|
|
|
pr0zac posted:This for instance demonstrates a pretty immature understanding of how a company functions as a whole. Seriously, go get lunch with one of those people sometime and ask them about their job and what they do all day. You'll likely find it interesting, have a better understanding for the value those people provide, and maybe learn a bit that will make you more effective in your job. 1. Somebody complaining that they went into the bathroom and discovered somebody using their stall. The gall! 2. Having to explain to a manager that his teammates were leaving because he was constantly yelling at them. That manager then yelled at her about that being impossible. 3. Making sure we didn't run over each other in the parking lot, which was little surprise to anybody.
|
|
#
¿
Aug 23, 2016 19:21
|
|
|
Has anybody dealt with the politics of an organization trying to take over the responsibility of doing every last ounce of design, architecture, framework definition, and infrastructure from all other software teams? Just to emphasize, I mean all that, but none of the coding. It seems like the dumbest thing in the universe, and I am pretty sure every road that leads that way is paved in a pile of corpses and ends in a bottomless pit. I understand the idea of architecture review boards and the like, but I'm dealing with a group that gets pissed if I declare an interface or something while solving a problem. Like, I have to literally code between their lines, and if I need that abstraction, I file a ticket to them. The thing is that they don't have a direct responsibility towards the work. They just are the people that are "here to help" with the infrastructure and blablabla I want to rip off my face.
|
|
#
¿
Aug 24, 2016 17:25
|
|
|
Necc0 posted:It could be worse. You could be working on a 20 year old code base with zero documentation, review, or QA where every engineer interprets the product manager's requirements however they like You think these people that want to do all this architecture stuff will give us something new? Some of them have been pushing the same poo poo for over a decade too.
|
|
#
¿
Aug 24, 2016 20:36
|
|
|
leper khan posted:I can sort of understand not wanting to let a junior dev declare whatever publicly visible methods on an interface, but the concept seems pretty crazy and not typical in my experience. To make things funnier, I can't even get them to tell me what they think a framework actually is. I used the textbook definition when I proposed one and got shat on for creating infrastructure or something. My own boss suggested I just call it a library, which we started doing. Actually--thinking about it--when I mentioned defining a framework for this to be able to mock the environment, I wonder if some of them though I was using the term "mocking" as in "making fun of." God drat it that is pathetic. Just to check my math here actually: I would say that a bunch of code you call is a library whereas a framework is the Soviet Russia counterpart: it calls you. That generally implies a certain amount of architecture and defining lines to color in for some standard. I'd like to think that the people that are working on the particular topic day in and day out are in a better position to define the framework. Edit: Not only are they not reviewing it, they are not writing any of the code, or doing anything else but "helping" by supplying a framework to us.
|
|
#
¿
Aug 24, 2016 21:34
|
|
|
rt4 posted:How does a junior dev get any experience without designing their own objects, receiving feedback from seniors, and seeing their design in action?
|
|
#
¿
Aug 24, 2016 21:37
|
|
|
The good news is that the people trying to pull this infrastructure stunt were moved into a different organization and have lost a lot of reach, but they are still trying. It means they can still slow everything down. They had one meeting to make the argument and tried to conclude we needed six more to work together. My new management did not know the game, but they haven't bought it so far. Any advice on Amazon interviews? My first experience was extremely hosed up, but an actual development team wants me on the phone soon.
|
|
#
¿
Aug 25, 2016 05:56
|
|
|
Well, I had my second Amazon interview, and I think that went to hell. The first question was: Given an array of numbers, return an array that is the product at i of the array, excluding i. Do it in linear time without using division. I couldn't get there. I tried various lookups and the like and still couldn't hit what they wanted. I looked it up online afterwards. It's simple enough but I don't know how something like that would occur to me unless I was fresh out of school, majoring in CS, and had done a bunch of brain bender courses. That's not the kind of poo poo I deal with anywhere. The second question was: Given a string, return the first character in the string that occurs only once. I murdered that one in five minutes, of which I was mostly just typing it out. It's just a lookup table mapping character to count and its most recent index. Then you just filter through the ones that have only a single count and take the one with the lowest index. There's some semantics for empty strings and crap, but that wasn't a concern. I feel like if I want to get into these big places at all that I have to just practice mind bender bullshit stuff for a few months. Nothing else. It's really annoying how much programming interviews any more seem to just focus on Big-O complexity cleverness. Maybe I'm a lost cause, but between the programming I did as a kid, the programming I did in school, and the programming I've done professionally, that kind of stuff has been extremely rare. I mean, to the point where it's not even worth worrying about it, and just refreshing when a particular situation comes up.
|
|
#
¿
Sep 1, 2016 18:16
|
|
|
return0 posted:Was this phone/skype, or face to face? It was phone with an online collaboration link. We were able to talk to each other and type into a shared text box. Good Will Hrunting posted:Input : [3, 4, 7, 1] I'm not sure what you all are getting at: Given your input set of [3, 4, 7, 1], the output they wanted would be [28, 21, 12, 84] I kinda-sorta see that in what you have, but I don't see how you're getting there--with the zeroes in particular. necrobobsledder posted:First problem the way you worded it seems like this is what they want (basically a modified foldr as mentioned above): For people following along, there's a Stack Overflow about the literal question I had: http://stackoverflow.com/questions/2680548/given-an-array-of-numbers-return-array-of-products-of-all-other-numbers-no-div You have to keep them from using division because otherwise you'd just get the product of all elements together, then divide by the element at each index to get each individual element's result. It looks like the trick for a 4-element array is to multiply down. I think this is what the interviewer wanted to see: code:
|
|
#
¿
Sep 1, 2016 20:19
|
|
|
Good Will Hrunting posted:Or something like that. I agree that this isn't really something I would have come up with had I not been practicing these types of questions for a month. I can't imagine how any of this has to do with a job adding a help button to Amazon devices.
|
|
#
¿
Sep 1, 2016 20:38
|
|
|
Stinky_Pete posted:Adding a button isn't a job Writing the help system behind it definitely is. Did people really think I meant literally adding a GUI control?
|
|
#
¿
Sep 2, 2016 05:44
|
|
|
I was reminded of the problem today and got rather agitated. My wife had brought up some little thought puzzle to me yesterday, but I didn't really have time to crunch on it. I tried it today and had a problem to it: There are 100 people, of which some are honest and some are liars. They've just finished shaking some number of hands in another room, and now want to tell how many are honest and how many are lying. You ask them, "With how many people did you shakes hands in there?" The answers are 0, 1, 2, 3..., 99 down the line. I did the same old thing of not panicking over having 100 things and scaled it down to 1, 2, 3, and so forth to get a feel for it. I then realized I was getting all kinds of possibilities and decided to look up the question online. There was a missing piece of information: The participants know who is honest and who is a liar. They only shake hands with honest people. I guess it's also worth noting that there's no double-counting. So with that extra bit of information, it's very easy to figure out. You can mess with it if you want--I won't tell. I realized then that the interview problem I was dealing with was something I would have casually done with that, and is actually one of my usual chants for walking into these: 1. If the possibilities are suffocating, scale it down to 1, 2, 3, .. and then maybe think about 100+. 2. If they're still irritated by whatever I'm doing, use a lookup table. The problem was that the interviewer immediately asked me if there was a way to precalculate things without getting into n^2 complexity. So I got stuck on lookup tables. So it's very important to communicate these drat problems well. He also did say two loops will always be n^2. I mean, always. I think that is absolutely false. If you try to iteratively implement many of the log n stuff that would normally be recursive, you'd wind up with loops like that, but the logic inside them is not revisiting everything across a square domain. I suppose you could make the argument that creating two arrays with products in them is a precalculation, but it sounded like he wanted to do lookup tables. Whatever. I remember of another one that really got me that is frustrating to still think about. I was finishing up college and was interviewing at Microsoft for what at the time would be the Vista kernel. I was interviewing with the manager and he gave me what I thought was a really straight-forward and stupid question: how do you remove an element somewhere in the middle of a single-linked list? After just doing the mechanics of it remove the element and repairing the list, he kept asking me over and over how to do it. We got to a whiteboard even. He never elaborated any further, and eventually we were just back at his desk, going over my resume. I figured out later I had lost the whole thing right there. What I assumed afterwards he wanted was how to remove the centermost element of a singly-linked list, and presumably without having to reiterate it. That's the old two-pointer thing: one pointer moves twice, the other moves once; when the first hits the end, the second hits the middle, which puts you right there. I was doing similar poo poo in high school. Well, I guess we saw how Vista went. Anyways, I did poo poo out the n^2 solution so they knew I had a pulse and got into some rather elaborate stuff, and I did absolutely ace the second question. I didn't just get immediately rejected like I did for my first Amazon interview. That was for a system engineering position where they just asked me about the output from the top command and a bunch of other Linux administrative trivia, based on me saying that I have developed and worked under Linux, but never had to administer it. So they could still be mulling it over--or the extended weekend took over.
|
|
#
¿
Sep 3, 2016 09:06
|
|
|
Doctor w-rw-rw- posted:I'm helping a friend out and mock-interviewing her tomorrow. I've got a couple in hand already, but it wouldn't hurt to ask - What are some good questions to ask a developer at a junior level? (~6 mo-1 year experience) They've had the sweet taste of industry so they're better suited with dealing with a certain amount of ambiguity and mediocrity. You can throw actual examples of stuff at them and start having discussions about how they'd even start to try to approach it. A favorite of mine is to lead somebody through setting up a pathfinder for a robotic arm. For the problem at hand, they really just need to create and navigate a graph data structure, and the robotic arm is just a distraction. The good ones don't get overwhelmed with that even after I tell them it just has to navigate linearly between coordinates.
|
|
#
¿
Sep 11, 2016 07:56
|
|
|
Doctor w-rw-rw- posted:I don't think I've heard that question before. Can you expand on that? It was something I had to deal with. They were hardcoding positions for the arm to go through in space. It would linearlly move from one to the next. Adding a new coordinate was not something users could do. The programmer had to create it, present a UI for it, and then hardcode transitions. I saw that and said, "gently caress that." The whole thing can be data-driven as a graph data structure. On top of that, the logic for dictating transitions can be encapsulated so you get a State Pattern where each state also happens to have 3d coordinates. You can go further and provide context to each state, which IIRC gives you a Visitor Pattern. It sounds stupid to do this for just coordinates, but it was a life saver for adding multiple sources and targets. When it started, it could only have one source and two targets. At the end, there could be as many as there was space for the arm to reach. The arm spends most of its time dormant, and is also supremely expensive, so getting milage out of it is a big deal. So I show them the problem with the arm having to go from x to y, and then try to get them to generalize it. I ask for how they would define an input format and then keep scaling to see how they accommodate the changes. I would expect to be giving these tweaks to a junior developer in cases where I could not just dump an architecture on them, so I want to see that we can communicate. To keep things in scope, I give them a function that will linearly move the tip of the robotic arm to one given coordinate, and say that all coordinates and transitions have been verified not to bonk anything. I then establish I want an engine to navigate and service coordinates.
|
|
#
¿
Sep 11, 2016 08:26
|
|
|
I guess we're talking about MVC framework stuff now. Perfect segue for me. Is Django still a well-regarded pick for a basic shopping-cart-style site? I'm talking about looking through lists of stuff, adding them to a personal queue, and then getting them in some way through some back end. I'm not talking actual retail. Since my place loves Python generally, it seemed like a very good choice in the general scheme of things. I'd like to do some nice client-side stuff too eventually, so I was looking up Node.js. I didn't really know squat about it, so I was surprised to see it basically has its own service and everything. I don't think I want to make a rich client application like just about any site ending in .io, so I wonder if it's overkill. In another rant, I mentioned unilaterally playing around in Django with a remote team in a call, and they started laughing about it as another one of those bloated, kitchen sink frameworks that are hard to use. I'm disgusted. Of course, their pile of PHP helpers is clearly superior.
|
|
#
¿
Sep 28, 2016 20:39
|
|
|
Yeah I had to deal with some butthurt for showing internal code in an antipattern talk that was two years old and written by somebody that I know had left the company.
|
|
#
¿
Oct 3, 2016 15:59
|
|
|
I would be more concerned the company is doing this "we are a startup!" thing when they have more than 200 people. I would wonder if they are really just a mid-size company using that as an excuse to get you working 80 hours a week.
|
|
#
¿
Nov 29, 2016 09:14
|
|
|
Is there a nice list of algorithm bullshit problems online that doesn't require signing into something? I have the prospect of a few interviews coming up. Previously, I seem to not have impressed people with finding the super-duper-most-efficient way of detecting anagrams or whatever and I need to brush up.
|
|
#
¿
Apr 25, 2018 14:52
|
|
|
Well one's Amazon and a previous interview for a different position had them gaming for something. I honestly never figured out what they wanted. It might have been using a hash lookup or something else. I don't know. So I was planning to just go through a few, brute force them for reference, and then contrive some bizarro, overoptimized crap.
|
|
#
¿
Apr 25, 2018 17:05
|
|
|
Blotto Skorzany posted:The flaws of the "data structures and algorithms class redux" approach that has been in vogue for a few years are way less onerous than this.
|
|
#
¿
Jun 7, 2018 21:57
|
|
|
My problem with take-home work before is they tend to assume it will take 30 minutes based on somebody that already knew the answer and just had to rattle it off. Maybe a "30 minute" thing would take that long if I happened to be sitting at my computer, environment up, everything quiet, and I'm ready to go. Maybe. But if you start the clock at that moment, then you're being really unrealistic. Then you factor in they're reading the thing for the first time right there, the who the hell knows.
|
|
#
¿
Jun 8, 2018 20:36
|
|
|
I feel like I need to get a data structures text book or whatever because I never took a formal data structures course in college. It was all just scattered across various computer science courses. The only smarts I've had for trying to answer interview questions has been that they probably want me to sort and then hash something to solve a problem. This breed of questions really isn't inspiring me in my job search. I have an Amazon one in the queue that I'm thinking of just burning up because I'm pretty sure it's just like this.
|
|
#
¿
Jun 15, 2018 16:28
|
|
|
ForrestPUMP69 posted:Their interviews seem more weighted toward behavioral questions, so you might as well go through with it. I had an onsite last year and it was a 60/40 split of behavioral/technical. The coding was mostly centered around BST's and wasn't terribly difficult. System design was tricky, but I did ok. Where I screwed up was the "leadership principles", which seemed like total BS and I didn't take seriously enough. Unfortunately I bombed the final interview with the team manager, as it was fully focused on these goofy principles. It's all started with phone screen stuff for me. This would be the third team trying to interview me at Amazon over the past few years. The first one had me do an interactive coding exercise to find how many different combinations of elements in a list sum to a certain number and like hell in 30 minutes or so could I figure out how to do it in faster than n-squared. The second one wasn't a big problem IMO. I had been a "system engineer" for something like a year-and-a-half. My job was solving integration problems between different software internally. The rest of the planet apparently believes it means "super-duper sysadmin" and you get a ton of really obscure Linux trivia questions. I had that problem at another place too. I just took that off my resume. Granted, I'm stuck in a Windows shop and I use Windows at home mostly because of Unity 3d, so my Linux skills have degraded, but they were interested in a lot inode fuckery and such.
|
|
#
¿
Jun 16, 2018 15:01
|
|
|
Yeah better than "I am an engineer! I only care about facts and am oblivious to natural logical biases and fallacies!"
|
|
#
¿
Jun 20, 2018 07:15
|
|
|
What Good Will Hrunting's situation sounds like to me is having a bunch of people litigating over the complete structural correctness of code being shat out in a text editor in real-time while a bunch of people are watching. He can't even sit back and think for himself every little bit to catch stuff because somebody's just peck, peck, pecking at it. It's the worst kind of paired programming. Edit: "He," I think.
|
|
#
¿
Jun 20, 2018 22:14
|
|
|
Good Will Hrunting posted:Today I interviewed at a place that seemed to really like me, and I'm pretty sure I nailed all 3 of the interviews. They checked pretty much all the boxes except one: a lot of their back-end code is Rails services and a Rails monolith-type app. Not all of it, but as far as I could gauge they were still in the process of shifting gears towards Scala micro-service architecture. The most senior person I interviewed with said that teams are structured so that each team is allowed the autonomy to choose their tech at this point and you're placed on a team based on how it fits you plus what you're interested in which is nice, but I'm still a bit concerned about making a decision if I do get an offer.
|
|
#
¿
Jun 21, 2018 22:16
|
|
|
Good Will Hrunting posted:I did though! In terms of performance and space constraints I did! Just not the best variable names and "structure" I guess! "I don't think I could go have a beer with this guy." (doesn't actually like beer)
|
|
#
¿
Jun 26, 2018 00:21
|
|
|
Good Will Hrunting posted:"Weak at overall system architecture" ... am I actually? I seem to be discussing trade-offs and design decisions for extremely loosely specced things, is it just that I'm not giving the answers you want to you question..?
|
|
#
¿
Jun 26, 2018 05:56
|
|
|
I have to give good props to Good Will Hrunting for staying in the game because this is the point where graybeards decide it's because they're not young and Indian and then go completely nativist. Funny enough, I was on a phone interview the other day with an engineer and we ended the call with interviewing horror stories. He apparently had to implement a fully-functional red-black tree on a whiteboard for an Amazon interview some years ago. They just expected he'd be able to do it. I can only imagine this being useful for an internal team making a container library. FWIW if I were manager the last thing I'd want is an internal team making containers. gently caress that. I have an Amazon interview in my queue and the recruiter outright said to study up and do some reviewing of algorithms and data structures. I'm just doodling on a few things I personally find interesting and enriching, but now I'm tempted to even stop that and just spike the interview. It looks like Amazon interviews are just trash now. A decade ago, I could expect to do something with pseudo-code that involved a dash of recursion.
|
|
#
¿
Jun 26, 2018 16:04
|
|
|
Good Will Hrunting posted:I don't even know how to get better at those kinds of interviews. Am I loving stupid? Do most people solve these in 35 minutes in a glorified Word document? I actually ran into a situation where a referral got me 99% of the way through an interview process, but the hiring manager went against everybody's recommendation because I didn't seem to have a good attitude. I had done the interview after my sabbatical got delayed over and over again due to politics. It had been about six months and I hadn't had a vacation because I had this potential sabbatical dangling in front of my like a carrot on a string on the end of a stick. I was worn the hell out. I aced all the questions but the manager just didn't like my vibe. My buddy was pretty pissed about it. The networking thing IMO works better if you're regularly going to conferences and such and have people there with which you're regularly catching up. Given how few and far between I get into intelligent conversations about computer software stuff, the whole concept is starting to warm up on me, but it's a lot of hustle.
|
|
#
¿
Jun 26, 2018 18:27
|
|
|
Good Will Hrunting posted:Personality is honestly probably my strongest point after communication and expression.
|
|
#
¿
Jun 26, 2018 20:47
|
|
|
Good Will Hrunting posted:It's too bad I can't code well! But then again, who does? Good news! Robots can't code either!
|
|
#
¿
Jun 27, 2018 01:26
|
|
|
We should just get on Discord and something and interview you and see what the hell is going on. Somebody can even play the part of a horrible interviewer and make you guess a number between one and a million or something.
|
|
#
¿
Jun 27, 2018 18:31
|
|
|
Good Will Hrunting posted:It's a job scheduler we use internally. I forgot what an anagram was once and wasn't getting any help from the guy over the phone. I guess they were a Serious Anagram Detector Company.
|
|
#
¿
Jun 28, 2018 15:40
|
|
|
Good Will Hrunting posted:So basically I am expected to come up with a flawless solution with zero bugs that I have the time to go through and test in the constraints of 30 minutes?
|
|
#
¿
Jun 28, 2018 22:13
|
|
|
Good Will Hrunting posted:Anyway, I made it to the final round with a place I'm interested in. Just 1 hour culture fit at an office event + 1 hour with the CTO left. I've had 4 tech rounds already, which apparently I passed regardless of not having slept properly all week as I've been mourning the loss of my poor doggo earlier this week which has been really tough on me. Not even sure how my brain is functioning, to be honest.
|
|
#
¿
Jul 12, 2018 17:52
|
|
|
|
| # ¿ Apr 26, 2024 23:21 |
|
|
JawnV6 posted:Intel explicitly encouraged folks to hop teams all the time. I wish they'd stay a little bit longer. They learn how to push a review to gerrit and not to use arbitrary string literals as comments but leave before they can learn how to resolve merge conflicts and write their own objects.
|
|
#
¿
Jul 13, 2018 03:52
|
|