Dear ML/DL People

Oracle ovKR85
Apr 12 38 Comments

If you are ML/DL practitioner and looking out for job, please say good but to you AI/ML/DL hard earned knowledge; As that holds no value in landing you a job. You should spend your years/month on LeetCode/HackerRank etc.

Hear from me, who has applied multiple AI/ML/DL techniques to solved real life problem, won many competitions against Phds, but have been not been able to get good enough calls from any good company and feel lost in space of LeetCode.

comments

Want to comment? LOG IN or SIGN UP
TOP 38 Comments
  • Uber ipl29
    Do you think communication might be an issue?
    Apr 12 1
  • Microsoft FucMatlab
    Sorry OP, considering that you don’t have a Ph.D that’s what’s preventing you from getting positions. Me, (A Ph.D level DL/ML Professional) has no issue landing high TC jobs. In fact I stopped having to apply for jobs back in 2017 because recruiters reach out to me.

    ML/DL is still a very “academic” field. The academic world is incredibly snobby. As a Ph.D who preferred to go into AI project management making a very good salary, even more so than the majority of my professors, I can say a hard nosed academic would look down at my job since I am not doing pure research.

    Even a Master’s in the field is not really “enough” as they will not be seen as golden as a Phd, UNLESS, you got your masters under a leader like Andrew NG, or Geoffrey Hinton.

    I say this not to “brag” but to give a realistic perspective. Even so, I value the foundational knowledge as well such as optimization, because DL is essentially a fad.

    When all you do is kaggle, and take a few coursera courses, you are a fad DL “practitioner”. When you go through school, take all the math, physics, probability, statistical, machine learning courses etc. You are a person with a breadth of knowledge who also happens to study ML/DL and thus you are flexible in your knowledge.
    Apr 12 23
    • New / R&D zorkan
      I think this is hubris to say DS teams cannot influence research teams. Everything in industry is influenced by value add over the axis of time. Good managers will shield you from it, but at the end of the day that's the name of the game. If you zoom out 5 years down the road and find that some research efforts are not adding any value (intangible: publications/patents/ect.., tangible: revenue/user engagement/ect...), those teams will likely be pushed into different directions. Eg; Apple self driving efforts probably included DL researchers.
      Apr 12
    • Microsoft FucMatlab
      I worked as a DL researcher in an Apple Special Projects group but since the project is not out yet, I don’t have anything to say in response.

      However, I will say it’s not hubris, it’s common sense. The field of DL research is primarily coming up with frameworks/models etc for projects that may not even be materialized yet. For example, thinking data science has any effect on quantum image processing research at MSR, or Google, is absurd. Also datascience is very easy, it’s part of the fundamentals in learning about deep learning. You can be a data scientist without deep learning experience but you can’t be a phd level DL researcher without understanding data science.
      Apr 12
    • New / R&D zorkan
      There is obviously variance in what the day to day looks like for a DL Researcher/Scientist. I've had opportunities to do it in academia, government, and industry (not at the tier of LeCun/Hinton --although I have had the pleasure of meeting Yann a few times) and the control you have over data/deployment/direction is always different. The takeaway is that it does not look like a Kaggle competition, sans some circles in academia.
      Apr 12
    • Microsoft FucMatlab
      I thought you said it did? Which is the position I was against.
      Apr 12
    • New / R&D zorkan
      "Yeah, there isn't much overlap between Kaggle and DL r&d in industry." --original comment. We are in agreement on the Kaggle point.
      Apr 12
  • Oracle ➕cowbell
    Communication is definitely an issue. I have no idea what this means:
    "If you are ML/DL practitioner and looking out for job, please say good but to you AI/ML/DL hard earned knowledge"
    Apr 12 0
  • New LC'ing
    Same story bro. I am non-CS Masters and I do ML/DL at work. I have a hard time getting interview calls. The experience itself doesn't get me calls. I am planning to go back to school to get a CS masters. My career has been at a standstill for some time now, and I am unable to get out if it.

    Unfortunately, I feel I wasted a few precious years of my life and a lot of $ getting a non-CS masters. All the predictive models I have built at work using regularization, elastic or neural nets apparently hold no value because I didn't study caching and I can't reverse a linked list. *crying* *crying buckets*
    Apr 12 0
  • Nvidia rosary
    Guys, a neutral network has gained consciousness and it's trying to communicate.
    Apr 12 1
    • Amazon leetN00b
      Almost passed the turing test.
      This one is less of a terminator and more of a marvin
      Apr 12
  • Amazon leetN00b
    "Please say good"
    good.

    I don't understand this post. You have to communicate your ideas and let the interviewer know your thought process.

    Listening to feedback also helps. The first response mentions communication and you respond by saying "I don't think so". Interviewers can provide hints when you are off track, if you ignore it and think you know better, that is another way to fail an interview.
    Apr 12 0
  • Fox Broadcasting vKlrt05
    Well, can you refer us to some of your publications!
    Apr 12 0
  • Oracle ovKR85
    OP
    No, I don’t think so. Its just leetcode..LeetCode..LeetCode
    Apr 12 0
  • Oracle / Eng wpooh
    Why is DS/ML diff from software engineers? I know the skilss are different but its not like software engineers do Leetcode kind of problem solving in our real jobs

    In all fairness job interview standards should be same for everyone 😂
    Apr 12 3
    • Microsoft FucMatlab
      It’s not, we still do the leetcode interviews plus domain interviews. The loops are harder than pure swe.
      Apr 12
    • Qualtrics / Sales
      '🈚🀄🎶🈶

      Qualtrics Sales

      PRE
      Microsoft, Qualtrics
      '🈚🀄🎶🈶more
      Harder loop, not much better pay
      Apr 12
    • Microsoft FucMatlab
      Yeah, I don’t know where you’re getting that from, we easily command 200k + upon graduation.
      Apr 12
  • Qualtrics / Sales
    '🈚🀄🎶🈶

    Qualtrics Sales

    PRE
    Microsoft, Qualtrics
    '🈚🀄🎶🈶more
    Try MLE positions, and yes whatever roles that require coding will test you on coding.
    Apr 12 0