Has anyone been bait-and-switched when applying to machine learning positions?

Microsoft Yiwb71
Aug 6, 2018 22 Comments

I work on ML infrastructure at Microsoft. Mostly accelerating common ML libraries. I applied to AWS for an ML position but when talking to one of the engineers he told me that his team doesn't really do any machine learning work -- they are just given the model from a research team. He said he and the whole team want to do more ML work but in reality they are just taking the model and building a service around it.

This seemed like a total bait-and-switch to me. I've read about how a lot of ML engineers spend most of their time doing data prep, but I hadn't heard of something like this.

comments

Want to comment? LOG IN or SIGN UP
TOP 22 Comments
  • Google / Eng GeoffDean
    It’s really common
    Aug 6, 2018 8
    • Microsoft Yiwb71
      OP
      Did you apply as an ML engineer or a software engineer?
      Aug 7, 2018
    • Oracle / Eng Lknbvgfd
      I would like to know the same.
      Aug 7, 2018
    • Google / Eng GeoffDean
      ML engineer, gives you at least title clarity, but I also know pure SWEs who do ML here. ML is all over... I’m sure plenty of folks who want to work on it don’t, but typically it’s easy to get put on projects that do ML if you let your manager know that’s your expectation.
      Aug 7, 2018
    • Microsoft Yiwb71
      OP
      Ok but what type of work is that? I'm guessing it's more data pipelines and infrastructure than actually developing a model from a dataset
      Aug 7, 2018
    • Google / Eng GeoffDean
      It’s basically end to end, though to say we have a lot of researchers who are the ones predominantly focused on model development and benchmark. With that said, ML isn’t just deciding how many convolutions layers, a lot of it is the actual system. We touch everything, though if you want to do only the model architecture part, you’d have to go for the research position.
      Aug 7, 2018
  • Amazon ladygaga
    Very common. You’re part of a ML team but all you do is software engineering
    Aug 6, 2018 0
  • Citibank TY3L4
    I do that as an ‘ML Engineer’
    Aug 6, 2018 0
  • Axtria 💗69
    Very common scenario
    Aug 6, 2018 0
  • Nvidia zVEC05
    I’ve been getting that a lot from Amazon & Google. They try to lure me in with a promise of ML; but then pair me with non ML or tangentially ML teams.
    Aug 6, 2018 0
  • Microsoft etrigan
    Typical scenario across several large enterprises claiming to be ML leaders. But it's not so surprising to me because if you think about it, these companies really can't start investing in employing full-time "data scientists", at least not yet. Data Scientists are already overrated and so are "ML Engineers". Doesn't stop you from unpacking the R or Python algorithm for your own learning. So enjoy the ride for the next couple of years before the real data scientists take over the world.
    Aug 6, 2018 0
  • Lockheed Martin / Eng
    Mxci11

    Lockheed Martin Eng

    PRE
    Booz Allen Hamilton
    Mxci11more
    Not exactly for ML, but I was hired as a Data Scientist but worked as a Data Engineer for my last job.
    Aug 7, 2018 1
    • Google poDV10
      Could you please describe the ‘Data Engineer’ role that you have to do? Maybe you had to do both because there aren’t dedicated data engineers?

      From what I know it’s a step up for a Data scientist to move onto Data Engineering.

      Moreover, Data engineering + Data scientist will be a killer combination. Once you have amassed real good experience on both sides, try out AirBnB, FB, Netflix.

      So if you are getting the opportunity to do the real ‘Data Engineering’ work, don’t think of it as bait and switch. Instead it’s a blessing in disguise. Make the best of it.
      Aug 19, 2018
  • Amazon kdagjd
    What role are you applying to? If it's SDE then it's very expected that you build a scalable service on top of a given model, because you're an SDE. But if you're a research scientist it's purely ML (mostly).
    I'm an SDE and our sister team is a mix of SDE and research scientists (mostly scientists and just a couple of SDEs). We build production scalable services on top of what they model. Their SDEs are mostly a bridge between an engineer and a scientist. They don't build scalable services, but they build prototypes for the scientists' models and they understand the ML as well. You might be interested in such roles.
    Aug 6, 2018 1
    • Microsoft Yiwb71
      OP
      Messaged :)
      Aug 6, 2018
  • Microsoft motiv8
    At Microsoft ML positions are mostly with Microsoft Research.
    Aug 6, 2018 1
    • Microsoft Yiwb71
      OP
      There are plenty that are not... but I don't want to reveal what team I'm on
      Aug 6, 2018
  • Microsoft compiler
    Yes Amazon tried to do it with me.
    Aug 7, 2018 0
  • Two Sigma PE2
    Because ML is a buzz word for recruiting. I regard recruiter emails about ML jobs as NEGATIVE signs. though I have done some applied ML work previously, I didn’t actively participate in this field in recent years. I don’t see why recruiter can easily discover my prior ML experience. So my conclusion is that’s a recruiting buzzword and I should avoid those jobs.
    Aug 6, 2018 0