Non technical data science manager

Seagate
scf768

Go to company page Seagate

scf768
Mar 10, 2021 105 Comments

I am a data scientist (Ph.D.) & report to a non-technical manager who has done MBA strategy who is in the company for a long time...

I have a hard time during my 1:1 & while explaining the project status during the meetings.

if I say this model is overfitting that person has no clue.. If I say I am spending time on data preparation the answer is you spend too long on nonproductive tasks. I am like bad data is gonna screw up... 🤦‍♀️

There is a problem I am working on for like ~500k rows & XGBoost works best but I get a comment like why not implement a neural network? 🙅‍♀️

My manager sees some youtube video accidentally once in a while & says why not try reinforcement learning for regression use-cases 🙄

How do I deal with this kind of situation? 💁‍♀️

I am curious to know how technical are managers managing data science teams in your organization...?

I am in this org for ~1.5 yrs

TC:180K #data #dataanalytics #datascience

comments

Want to comment? LOG IN or SIGN UP
TOP 105 Comments
  • Change your job. In Data Science, never work for a manager who has no background in Data science, else you will always be underrated. For a short term fix, as many suggested try to explain things in simpler terms without using too many jargons.
    Mar 10, 2021 6
    • Google
      __human__

      Go to company page Google

      __human__
      The chain inevitably will report into someone who has no clue eventually, but that should happen high up, not at the IC level.

      The chief data scientist (or whoever is in charge of the data science org) should be more technically skilled than this person is describing, and should be able to shield their data scientists from the noise of stakeholders trying to interject uninformed opinions into actual analysis.
      Apr 6, 2021
    • Robin
      Loop_hero

      Robin

      Loop_hero
      Lots of good items in the thread. Also, it is good to get buy in from a stakeholder or someone to act as a translator. If the manager is clueless, see if you can pull your stakeholder in to help communicate the concepts.
      Apr 8, 2021
  • Microsoft / Data
    CBso70

    Go to company page Microsoft Data

    CBso70
    Quickly implement a neural network and prove him wrong
    Mar 10, 2021 8
  • Talk to your colleagues/ look at their work. How technical are they being with their work?
    In the short term, explain your logic to manager once. But if they don't like it, do what they say. Interview and get out of there
    Mar 10, 2021 4
    • Uber / R&D
      babygrogu

      Go to company page Uber R&D

      PRE
      Uber
      babygrogu
      +1 to advice from DD. This is an excellent opportunity to work on your communication skills to nontechnical stakeholder. Full honesty not always reqd
      Apr 17, 2021
    • While honing your business skills, you will lose or at least get behing on your technical skills. Jack of all trades, master of none. Be aware of the effect and decide what you want.
      Apr 19, 2021
  • Lyft
    Road Head

    Go to company page Lyft

    Road Head
    1:1 with your manager should focus high level and on your career path. Don’t go into technical details or undersell yourself. Can “data preparation” be rephrased as “feature engineering”? Or can you consider it part of the modeling process and say you are working on modeling?
    Mar 10, 2021 2
    • Seagate
      scf768

      Go to company page Seagate

      scf768
      OP
      Yea 1:1 is basically starting with what are you working on... so you know what I mean they never shift the topic on goals / careers
      Mar 11, 2021
    • AMD / R&D
      EPYC'er

      Go to company page AMD R&D

      EPYC'er
      You are responsible for your career. Send out an agenda before your 1-1, and clearly call out what you want to discuss. Ask specific, pointed questions about your career path. This is your responsibility, not your managers.
      Apr 10, 2021
  • eBay
    wdCe67

    Go to company page eBay

    wdCe67
    please interview and change ur company or group as soon as possible. I had a very similar experience where my manager had never even developed a regression model. She would always told me that science models work randomly so you should not spend a long time developing that. or she would ask us to replace xgboost with rule based filtering. At the end of the day, she wrote a horrible performance review for me since she had no idea what i was doing and couldn't get steps needed to develop a model. While looking for a new place, in the short term just use lots of visualization stuff
    Mar 11, 2021 4