Intel interviewers knew less than me

Nvidia Obgr
Mar 23 51 Comments

The position was for a math library. The interviewers have never heard about an algorithm to calculate Chebyshev series. It's a very basic stuff for function approximation. They didn't know about the Ryu algorithm. During my presentation one guy asked if a least square task is ill-posed. It was for orthogonal functions, of course it is not.

They rejected me. :)

comments

Want to comment? LOG IN or SIGN UP
TOP 51 Comments
  • Amazon tofu bird
    Now you know the reason why they rejected you lol
    Mar 23 0
  • Amazon Algoo111
    It is for the better. You probably wouldn’t have been happy if you have to explain these algorithms to everyone on the team on a daily basis. I guess if you have loved that you would have gone into teaching...
    Mar 23 1
    • Nvidia Obgr
      OP
      Absolutely. I do not regret. The last interviewer was the manager. Gosh, I wouldn't want to work with him. It was after 8 hours of interviews. Me like: you can't do it, you need two multiplications. He: I don't want to use multiplications. Me (after thinking for a while where is the trick): I give up. He proudly writes down a solution that has two multiplications. Me: hey, that's what I told you. He then blames me I didn't understand him correctly.
      Mar 23
  • Yelp / Eng sJJv80
    Jesus Christ. I had a grand total of 1 numerical analysis class years ago, and I know about Chebyshev series. Never heard of Ryu algorithm, but, if you could competently explain it to me, I’d rate that highly.
    Mar 23 5
    • Nvidia Obgr
      OP
      To be honest they knew about Chebyshev series. They did not know a numerically stable algorithm to calculate them. It seems like at least for sin(x) they calculate the coefficients in front of powers of x and then use the Horner method. Which requires one more power term to achieve the same precision. The Ryu is quite new, but already widely accepted, algorithm to convert floating point numbers to strings. It was developed by Google. Microsoft has already implemented it in their standard library.
      Mar 23
    • Yelp / Eng sJJv80
      Still, assuming these guys are PhDs in numerical analysis or, at least, longtime practitioners, they should know the barycentric interpolation algorithm is fast and numerically stable.

      I mean, I didn’t know that, but it was the top result of a 5 second Google search I just did, and it came with a reference to a proof of stability.
      Mar 23
    • Nvidia Obgr
      OP
      I mean the algorithm to calculate the Chebyshev series, the actual sum itself, if you know the expansion. The trick is to start from the end and use the recurrent formula to avoid calculating the Chebyshev polynomials explicitly.
      Mar 23
    • Yelp / Eng sJJv80
      Oh, yeah. They taught me that in undergrad, lol. It’s similar to calculating Lagrange polynomials, IIRC. I thought you meant evaluating functions expressed in Chebyschev basis form. (Those subtractions in the denominators of the barycentric formula do look scary, and the proof of stability is moderately recent, but I’d still expect anyone using Chebyschev polynomials to know that.)
      Mar 23
    • Nvidia Obgr
      OP
      Great, thank you for confirming I am not being surprised by the lack of some exotic knowledge in Intel. It's a lack of basic knowledge. ☺️
      Mar 23
  • Intel / Eng mr_pleb
    Why did you interview at Intel in the first place if you have high aspirations ?
    Mar 23 2
    • Nvidia Obgr
      OP
      Because the position seemed interesting.
      Mar 23
    • Samsung EternalACT
      Working on something esoteric is a death sentence at Intel.
      Mar 23
  • Amazon gw
    Classic case of b players hiring c players.
    Mar 23 1
    • Google camelCase
      Typical of big old companies.
      Mar 23
  • Intel exceed
    Intel ppl think they are smart
    Mar 23 1
    • Intel dObv43
      No they know they are dumb they are just scared to be exposed
      Apr 5
  • Nvidia bdei51
    Whyd you consider going from Nvidia to Intel?
    Mar 23 5
    • Nvidia Obgr
      OP
      It's actually two questions. Why from NVIDIA? If you work here, you know the internal problems. My personal story is work 60—80 hours a week, i.e. including weekends and holidays. No appreciation for hard work. Micromanagement. Favoritism. Low pay. Not fully using my brain potential. Why to Intel? I don't have any particular wish to work there. It's just the position looked interesting.
      Mar 23
    • Samsung EternalACT
      With the possible exception of work hours, I guarantee all your objections are far worse at Intel.
      Mar 23
    • Nvidia Obgr
      OP
      Good to know. Thanks!
      Mar 23
    • Nvidia DE2019
      60-80 hrs a week and no appreciation ..it sucks big time . sorry man
      Mar 23
    • Qualcomm / Eng hjklg
      Does that mean tc at Nvidia is 1.5x -2x more compared to intel? If we add OT factor itll be something like 2.25x-3x more?
      Mar 24
  • Intel DduE35
    You totally dodged a bullet. I mostly interview internal candidates. They all have graduate degrees and 7-10 years of experience. Few of them can come up with an algorithm for basic string manipulation or basic bitwise operations let alone code it up. The other one percent just study leetcode answers and regurgitate those in an interview even if it doesn't actually solve the problem.

    They probably didn't reject you. They were thinking of your future prospects knowing how bad it is at INTEL.
    Mar 24 1
    • Samsung EternalACT
      Intel is fine with hiring smart people if they can be abused and dominated. A lot of the older engineers still at Intel are just too meek to recognize that management and their pet engineers are raking it in at their expense.
      Mar 24
  • The JD didn't provide any hint of the team? This is in Santa Clara? Also what grade was it for? We could give you the salary band for this position, which undoubtedly would pay you less than Nvidia does before taking into account work hours. If the team is as incompetent as you say it is, which wouldn't surprise me, then you could probably get away with working 30hrs/week and getting great performance reviews.

    In the end, you wouldn't want to downgrade your employer to Intel so the hiring manager did you a favor. Intel has been doing the C players hiring D players thing since the mid 2000s from what I hear -- I came from an acquisition a few years back and I think this place is probably the worst place I've worked in my life (startups to other well known tech companies). Just take a look at the target schools Intel hires from: Portland State, ASU, etc. I'm only cruising here because of some family circumstances, but will GTFO as soon as I can.
    Mar 25 5
    • Nvidia DE2019
      Come on ASU CS is in top 40 .
      Mar 25
    • Back in my day, Nvidia used to target mostly Stanford and Berkeley grads in the Bay Area, and you had to have a GPA > 3.5. I guess not true any more?
      Mar 25
    • Intel itlsux
      Even ASU grads are rare now days. Its mostly oregon state, psu, and other non ranked schools like kansas state, montatana state
      Mar 25
    • Maybe in Oregon then? In SC/SJ, in the past couple of years, I've met a handful from MIT, Stanford, Berkeley, UMich, even Caltech of all places. But they were mostly from acquisitions or in "high impact" groups. And mostly they are not happy and want to get out if they haven't already. The few that are somewhat content are G9 and hoping to get that G10 Principal Engineer nod soon (dunno if that will happen).
      Mar 25
    • Nvidia Obgr
      OP
      I don't regret they didn't take me. I posted this just to share the experience, not because I am frustrated. The manager showed himself as a bad communicator, I would have rejected the offer even if I got it.
      Mar 25
  • Intel Grtv08
    They are going down for sure. They don’t know how to retain talent. Politics there is at peaks.
    Mar 30 0
  • Intel curious_ee
    Haha congratulations on getting the reject! You deserve better and they deserve worse :P
    Mar 26 0
  • Intel RiGD34
    Yes they rejected you because you knew way more than them now go find your peers who understand what you are talking about.
    Mar 24 0
  • Aurora bystander!
    Apply to Aurora, we will adequately oblige your math background.
    Mar 23 3
    • Nvidia Obgr
      OP
      I checked LinkedIn for an Aurora company. There are several dozens with this name. Which one you are?
      Mar 23
    • Nvidia Obgr
      OP
      I have lidars experience. I applied to a position that asks for it. Check it out.
      Mar 25
  • MathWorks p8Rz2u
    How did you behave/answer when they didn't know something or asked a "stupid" question? Wouldn't be surprised if they rejected you due to behavioral reasons...
    Mar 23 1
    • Nvidia Obgr
      OP
      I wouldn't assume that, but if that were the case, I wouldn't be surprised as well.
      Mar 23
  • Intel dObv43
    I work at Intel..I work for a day and wait for a month for the team to catch up. U will decay here brother! You are lucky to be rejected, greater opportunities are waiting for you
    Apr 5 0
  • Nvidia DE2019
    🤣😂😂
    Mar 23 0
  • Qualcomm / Eng hjklg
    Interesting, was one of your questions how you would implement sin(x)? For the least square task, what do you mean by orthogonal functions? Curious to hear what other type of questions these roles ask for.
    Mar 23 4
    • Nvidia Obgr
      OP
      Oh no. The questions were pretty simple. Like solve quadratic equation. I already knew what they expect: don't forget to check if "a" is zero. If you really want to surprise numeric dudes, who assume you are an idiot, mention that you should never use (-b+sqrt(b^2-4ac))/2a. I have already knew that. Silly tricky questions. They are managed to show what you don't know, instead of letting a candidate to open up and show what (s)he does know. I talked about Chebyshev polynomials during my presentation. I tried to communicate with the audience during the presentation, and talked with them after. I also asked questions during interviews. I found that they were clueless, including the manager. sin(x) is my black box investigation of Intel's implementation. I know they use a polynomial interpolation and a Horner algorithm.
      Mar 23
    • Nvidia Obgr
      OP
      The least squares was to find a Chebyshev series coefficients using least squares. Least squares can be ill-posed, if you have an overdetermined system. Like trying to fit a 2D plane in 3D space on points that are located along a straight line. Since Chebyshev polynomials are orthogonal, all parameters are linearly independent and the task cannot be ill-posed.
      Mar 23
    • Qualcomm / Eng hjklg
      Does Ill posed in this context mean whether it's mathematically defined or there could be numerical stability issues?
      Mar 24
    • Nvidia Obgr
      OP
      The definition of an ill-posed task is if one of the following is not true: (i) a solution exists for any input; (ii) the solution is single; (iii) the solution is stable. Yes, if a problem is ill-posed it displays numerical instability. For example, numerical derivative is an ill-posed problem. A linear equation with eigenvalues close to zero is ill-posed. Fitting a plane on points located along a line is ill-posed. The challenge with ill-posed tasks is that a small variation in the input produces a huge response in the solution. This is where the numerical instability stems from. It usually requires some type of regularization.
      Mar 24
  • Intel WwfA21
    Were they fishing for information???
    Mar 30 0
  • Sorry about the interview. Sounds like you're into similar stuff as me. There doesn't seem to be many high TC positions involving more mathy type stuff.
    Mar 23 0
  • Facebook 😝oops
    Same happened with me in lyft!
    Mar 23 0