How to differentiate yourself as a Data Scientist?

dunnhumby / Data
ladka_

Go to company page dunnhumby Data

ladka_
Aug 13, 2021 3 Comments

Title is pretty self explanatory but adding some points to remove obvious answers. As for me, I am like any other average data scientist who can read blogs/watch youtube videos (without understanding the mathematics behind solutions/algorithms) and throw deep learning at any problem at hand ( also can crack interviews by cramming obvious ML questions) for justifying why should they get paid. Question is how can someone like me separate themselves from herd of data scientists within an organisation (pick any company name).

Please don’t mention following solutions:

1. Kaggle - As I said, average data scientist, not capable of becoming a grand master(other than that kaggle doesn’t count much)

2. Some other places which became too generic now a days:
a. Linkedin posts using copied code of random computer vision applications.
b. Writing blogs (medium etc.)
c. Writing data science leader in linkedin with udemy certifications in 2021.

#datascience #data #dataanalytics #datascientist #dataanalyst #machinelearning

comments

Want to comment? LOG IN or SIGN UP
TOP 3 Comments
  • PayPal
    periscope1

    Go to company page PayPal

    periscope1
    I am interested to what others have to say as well.

    I guess you can do
    1) open source contributions
    2) Papers, patents and packages
    Aug 13, 2021 1
    • dunnhumby / Data
      ladka_

      Go to company page dunnhumby Data

      ladka_
      OP
      I agree with your answer but there are two problems(from my point of view):
      1. It assumes you are not average (DS paper market is too saturated/cutting edge)

      2. Current office hours allows you to do these things.
      Aug 13, 2021
  • Target / Data
    etc42

    Go to company page Target Data

    PRE
    Cargill
    etc42
    To be separated from the herd, you need to learn the math, stat and optimization of DS/ML. That is the critical difference between your average and high performing DS.

    You can cram the ML questions, but any decent interviewer will be able to see through it quite quickly. Don’t get me wrong, you will still get hired and make good money, but you won’t be able to differentiate yourself.

    For eg. Everyone knows hyperparameter tuning. Can you explain to me how learning rates can be calculated for a simple quadratic problem?

    I personally don’t value candidates with a lot of kaggle/online database projects (of-course there are exceptions). Simply put, any kaggle problem has hundreds if not thousands of attempts. In most cases, it is really difficult to stand out in such cases, even if you are extra ordinary. As interviewers, it’s quite difficult to keep track of all these online certifications.

    On the flip side, if you show me even a simple project where you collected data through experiments or something similar, it shows that you have the ability to work on real problems.

    To summarize. Math over coding. Real projects over online challenges. My 2 cents.

    TC 200k Midwest. YOE 2
    Dec 18, 2021 0