Hacker News Reader

Comment by hintymad | Hacker News Reader

A side topic: how can a company find people like the author, who can really articulate how to build a system? I've been interviewing senior engineers for years, and more than 95% of them, if not more, are really box drawers. They throw terms around confidently but fail miserably the moment I ask for specifics. No, I don't ask for details like how Raft works. That level of details probably would fail all the candidates. Just simple things like how you organize the data, how you route the traffic, how you manage metadata, or how your data path keeps metadata in sync. Really basic ones, yet most interviewers fail. Indeed, the more senior they are, the more likely they lose touch of the details -- even those who claim to work on distributed databases.

View on HN

I recently interviewed with a company, and the interviewer asked me "how do you organize data." I wasn't sure if they wanted to talk about classes, modules, databases, k-v stores, hashing data and routing distributed requests to the same pods. I asked, and they answered "I mean in general, how do you organize data."

After talking for a bit about pretty much all of the above, the interviewer asked "have you used dictionaries?"

The reason I'm telling the story is, if a lot of your candidates fail to answer your questions, the problem might be in the question.

I would have halted for a moment and asked something like: organize data for what purpose? At which point I would expect there to be some amount of clarification. As it sounds to me like they're talking about organizing some set of data for lookup, as opposed to organizing data around how it flows through an application, for example

I often find when you are given questions like that no matter how you probe you get a really defensive interviewer who doesn’t want to give away the answer they are expecting. I’ve been in very similar situations with vague questions and I’ve tried to probe for more details and just been met with defensiveness with an attitude that says they expect me to know exactly what they mean

It is worse: your attempt to clarify the question may be regarded by some interviewers as a sign you are not senior dev (they think [erroneously in my opinion] that "senior" means that you must figure out fine details yourself).

This seems odd to me, though its entirely possible my experience thus far hasn't been this. I've been told at the last 3 places I worked that one of the reasons I was hired was how inquisitive I was during the interview stage, asking lots of questions before coming up with solutions

I am most definitely not ruling out what you're saying may be common case and I am just very lucky, but I do want to act as a sort of counter point to see if others can weigh in so we can get more of an industry sense around this

Anecdatum: my experience has been the same as others - asking questions about the interview questions generally leads to a bad outcome. In 25 years, I reckon it's less than a handful of times where asking questions of the interviewers has actually got a positive response.

Hell, even when in jobs, asking questions about projects I'm assigned has sometimes got a negative response...

Seems weird to me that asking a specific question about what it is you're trying to achieve would be negative. Especially since most interviews often open with feel free to ask clarifying questions or something along those lines.

If the prompt is unclear its worth getting clarity, just like if something is unclear in the job you seek clarity, I feel like people not asking questions would be a big red flag.

of course, asking too many (this is subjective but I think we can all think of a reasonable situation where there were too many questions being asked relative to their value) could be a red flag

> just like if something is unclear in the job you seek clarity

There are a distressing number of managers / leads I have encountered who consider that if you're asking for clarity, you're impugning their powers of explanation because CLEARLY they explained it well enough (after all, they understand it!) and you're either an idiot or being sarcastic to undermine them.

Could it be because the clarification questions are perceived as not understanding? Would it help if one prefaced with something like "Well, there's many different ways like X, Y and Z and the best one depends on the details of the requirements. Did you have something specific in mind and if so what are those requirements?"

By mentioning a few ways you'd show you're aware of various solutions, and you'd also provide context for why you're asking probing questions.

Then again, I haven't interviewed in quite a while (love my current job) so...

> Could it be because the clarification questions are perceived as not understanding?

Sure but taking that as "the asker is an idiot" rather than "I may not have explained this well" is all too common (I know I've been guilty of this more than a handful of times.)

> Would it help if one prefaced with something like [...]

I think if you're having to carefully phrase your (reasonable, obvs.) questions in order to avoid upsetting the interviewer, that's a bit of a red flag, no?

When I interviewed for software development, I used to have a very open-ended question about optimizing a system I would describe, and there was no right answer, there were tons of directions to go in, and the point was to get the interviewee to ask questions so I could figure out if they'd been around the block and which blocks they had been around.

Why would you only want cookie-cutter "correct" answers to your questions? That doesn't tell you shit about the candidate?

One a rare occasion I actually got feedback after the interview (it was for a startup), It was because I asked 'too many questions' it was decided that I'd need way more mentoring than they can afford for the senior position I was applying for.

That's infuriating. There's a clear distinction between asking "how do I do this thing?" and "what do you want me to do?"

I can't read your mind, and that goes both for interview questions and determining product needs.

There are plenty of different types of dictionaries, organizing their data differently.

I have had that exact experience before. It is fine, you realise the company is toasted and you move on (the place that asked me this had just got rinsed by an offshore consulting firm, ecomm site built with a graph database, comedic stuff).

> After talking for a bit about pretty much all of the above, the interviewer asked "have you used dictionaries?"

Hahaha classic.

I was once in an interview (which I failed) and was asked a problem related to some sort of monotonic queue. I wrote the solution and was going through it, when the interviewer asked: 'what CS concept are you using in this solution'? I didn't really understand what he was asking for and in turn I asked for more clarification like 'Are you referring to the data structures I used? Or the technique (it was some sort of greedy)?' 'no'. After a couple of back-and-forth questions and answers, he finally tells me he was looking for me to say 'a state machine'. Seriously?

"I was once in an interview (which I failed)"

You didn't fail that interview. They did.

Do you understand the difference between writing a blog post about something and being able to talk about it in an interview? The majority of work in software is not being an "expert" on any one thing because the one thing that employers require is changing quite frequently. Rather it is being able to build up experience and carrying knowledge forward to new areas.

Most technical interviews are conducted in a completely inane way where the goal is to memorize pieces of information. If that was how software worked, the best engineer would just be Google. 95% of people who know enough to build this system would likely be unable to answer your questions because, more than likely, they do not currently work in a job that requires them to use this information every day. This does not mean they do not understand it or are unable to build it.

The reason why companies can't find people like this is a combination of not understanding what software development is (and not understanding people, it is really the blind leading the blind but technical people are often as bad as interviewing).

You are asking completely wrong questions. You aren't starting from the right place.

> Rather it is being able to build up experience and carrying knowledge forward to new areas.

Then I'd expect the candidate could articulate her experience and knowledge in some way, no? Otherwise, how would I know the candidate has the built-up expertise? Of course, I assume we can only have interviews. Otherwise, we can have other means, like mini-project, onsite project, or a writeup. Some candidates do like the alternatives more, and some not.

> where the goal is to memorize pieces of information

I disagree. The goal is to see if a candidate does have what the candidate herself claims to know. I find it hard to imagine that a candidate claims to be an expert in a field yet couldn't articulate even one thing in depth for hours if not days, let alone 30 minutes of interview time. Note this is not about any specific details, but the general picture and insights that an expert can convey. This is like a PhD oral defense. The candidate talks about the topic that the candidate is familiar with, and the professors dive in on such topics. I don't see what's wrong with that.

Candidates aren't defending a PHd. The interviewer is nowhere near competent enough for this.

> The goal is to see if a candidate does have what the candidate herself claims to know.

The process you are using does not tell you that. You will notice that your explanation is full of things that you expect, not based in an understanding of how reality actually works. And the result is, unsurprisingly, one where you do not understand the output.

Maybe you're the problem? Looking at your questions, you're asking for opinions, not answers or solutions. Putting people on the spot is bad form in my opinion. If I have a technical challenge I want you to solve, I'll give you the time for that so I can actually see your work and discuss it with you.

I've seen people that can talk the talk but not walk the walk and vice versa, they can't articulate well, but their approach and work speaks for itself.

Maybe I am. For a senior enough role, my questions are open ended. I'd expect the candidate to drive conversations and examine alternatives. I ask follow-up questions based on what the candidate mentions. Yeah, such may be indeed their opinions, but at least they should be opinions derived from assumptions, constraints, and fundamentals.

That’s a terrible way to interview. Interviews are high stress, time constrained settings where you’re talking to people you don’t know. If you want thoughtful answers, you need to be precise in what you’re asking. You can’t be like “tell me about yourself” and then complain I didn’t get from the candidate whether steak is their favorite food.

I don't think that it is a terrible way to interview to all, for a senior position.

A generic question like "how do you organise data in this situation" is open-ended enough that it merits followups that the interviewee has to drive. If I was being interviewed, I'd ask "what are the operations you want on the data (random lookup vs scan, append, lifo access etc), what performance guarantees you want on those operations, what persistence and distributed fault-tolerance guarantees if any, etc.

It is no different from having to be in front of a client and teasing out what the client really wants. The interviewee has to guide the questions and answer them in a satisfactory way. Reminds of Prof. Manuel Blum (then at Berkeley), whose class project would be "think of a problem and solve it, and you'll be graded on both the quality of the problem and the solution".

I am not disagreeing with what the question is supposed to do and how candidates are supposed to answer.

The problem lies in the fact that it creates too much ambiguity in terms of how the candidate is evaluated. Some interviewers may give you credit for your approach, but like the parent, more often than not, you’re evaluated on getting to a specific question or direction, in a specific amount of time. And that just creates a bias towards you hiring people who have a chain of thought exactly like you, which may tbh be not correct chain of thought at all.

In a f*king database like everyone else, are you happy now? What have you learned?

You seem confident you wouldn't have failed the author of this blog post, but I wouldn't be so certain.

Interviews are a high pressure situation where people want an immediate answer, without much of any time to think, let alone research options or run an experiment. You get people's absolute worst possible results.

Where do you find the jobs where people actually care about engineering? Because real work is mostly careful plumbing without breaking what works, but in interviews you must leetcode. Real systems only work when kept reasonably simple, but system design interviews are all about drawing boxes.

I have worked in distributed databases. I can't seem to find any host organization that's actually building anything and really cares about those details. I know several other senior systems engineers in the same boat. where are you guys hiding?

I think there are plenty of companies building stuff like this, and hiring for it. The problem is going to be that the founding engineers have already hammered down (or attempted to) the bulk of the initial interesting engineering problems and what they want is foot soldiers to complete, fix, patch up, and maintain their design.

That's ok, because that's what the software engineering profession is, on the whole, for better or for worse.

But... the interview process should be clear about that, and the questions reflect that. Rather than "how would you build a high performance distributed database" the questions in those cases need to revolve around identifying problems in a design and how benchmarking, diagnostics, bug fixing, etc. would go. Even at the height of VC fever a couple years ago, being blessed with "Hey, here's some $$, go write a [database|operating system|game engine|programming language|other sexy thing]" doesn't happen really (and probably for good reasons.)

So, yeah, GP commenter is a bit disingenuous. There really are only a scant number of jobs which would involve actually being able to do these things, let alone answer how to do them in interview on demand. The jobs that are there are fixing someone else's thing where already they did this stuff.

Said interviewer would likely find a candidate who can answer all those questions... and then hire them to go write some microservices in C# for an insurance company, or join an on-call shift diagnosing production problems with an off the shelf database replication process or something.

Of course, not telling you what you don't already know, just ranting :-)

You find them via their blogs.

Try to interview more senior people, who still remember their C course in an University, or even know what is assembler. Currently, most graduates (from my experience) don't know how an integer is stored in memory.

I don't even know what a good answer to that is. Some integers are stored in registers. Some are stored on the stack. Some are stored on the heap. Or you could be asking how integers are represented, with two's complement being the usual way. Did I pass your ambiguous question? Probably not.

Reply on news.ycombinator.com