MLE role prep for New Grad

New kRgw75
Sep 8 14 Comments

Preparing towards a MLE role as a new grad, is doing Andrew Ng ML class, fastai DL class and some 10-20 Kaggle sufficient for the ML aspect of the interview?

Aiming tier 2 entry level MLE

comments

Want to comment? LOG IN or SIGN UP
TOP 14 Comments
  • New AWXh63
    Honestly courses, online classes, kaggle, and boot camps don’t really cut it. You would be better off making a blog where you find a random dataset and make a conclusion- ie cluster McDonald’s food using k means.
    Sep 8 4
    • New kRgw75
      OP
      How are ML technical interview questions for MLE like?

      Is the blog post suggestion a way to prepare for interview questions (going through the entire ML cycle) apart from being portfolio material ?
      Sep 8
    • New AWXh63
      YMMV but usually the stuff I made helped the most in tech interviews because I remembered it and could apply it to different problems. Otherwise you are trying to study for a question you don’t know.
      Sep 8
    • New kRgw75
      OP
      Is it common for them to pass u a computer, a dataset and a problem and watch you write panda numpy scikit code to process data, train several models and come up with a visualization and a few insights?
      Sep 8
    • New AWXh63
      Never been asked to design a whole system onsite. Usually the give you a homework problem like that to do before or after. For some reason it usually deals with real estate. Go figure. The other replies give a good overview of the type of questions you will get. Sometimes I get a lot of rapid fire questions like what is an eigenvector.
      Sep 9
  • AllianceBernstein / Data JhYt70
    Are you undergrad?
    Sep 8 8
    • New / R&D
      leti

      New R&D

      BIO
      first year phd student at unc cs department
      letimore
      expect at most leetcode medium.
      some popular questions:
      vc dilemma
      svm, pca derivation
      random forests
      ... specific questions regarding the role
      expect to hear their current issues as a problem asked to you

      if you have time, read Bishop's ml book. it is awesome!
      Sep 8
    • New kRgw75
      OP
      VC dilemma is referring to bias-variance?
      Sep 8
    • Google nanithefu
      I think he/she means VC dimension in computational learning theory?
      Sep 8
    • New / R&D
      leti

      New R&D

      BIO
      first year phd student at unc cs department
      letimore
      yes sorry about the mistake but they are trying to measure the same
      Sep 8
    • Google / Eng fsck
      Why the fuck would an MLE (as opposed to researcher) need to know about VC dimension? Besides a brief cursory exposure that it exists. This is a theoretical construct, used to establish some theoretical bounds of classifiers performance which are way too loose in practice for anything besides publishing papers
      Sep 10