This data looks at when data scientists start running heavy computation processes throughout the week over the month of February 2025.

Made with Python, Pandas, and Seaborn. The data used is collected from https://meerkatio.com, a VS Code extension for data scientists that monitors code execution to trigger notifications. MeerkatIO does not log user data so all notifications are in UTC time and with users all over the world I did not try to localize the timezones, although that would also be an interesting plot.

Posted by dajmillz

11 comments

  1. This data looks at when data scientists start running heavy computation processes throughout the week over the month of February 2025.

    Made with Python, Pandas, and Seaborn. The data used is collected from [https://meerkatio.com](https://meerkatio.com/), a VS Code extension for data scientists that monitors code execution to trigger notifications. MeerkatIO does not log user data so all notifications are in UTC time and with users all over the world I did not try to localize the timezones, although that would also be an interesting plot.

  3. Before I saw the explanation I thought you mean by “heavy lifting” is doing weights at a gym or something.

    And the only explanation I can come up in my heads how us Nerdy Data Scientists do any sort of physical exercise if if the bluest of the frequency means 10 pumps on a 5 pound weight or something.

  4. 3pm seems to show when the adderall begins to wear off.

  5. I’m not really all that surprised. I’ve definitely noticed that my productivity goes heavily down after around 1pm or lunch time (even if I don’t eat lunch). After lunch nap is real. I also notice how productivity goes down after Wednesday and looks to be the bare minimum on friday. All this does is convince me even more of a 4 day work week with less hours (or more flexible hours, like always be here between 11 and 3 but the place is open from 7-6 or something).

    I would bet most of the drop off we see is when a large amount of people basically stop doing work for the day, or just do the bare minimum. Can some data science work be automated and thus part of the drop off is just the computers working without the need for constant input? I am genuinely asking.

    cool chart, would love to see this repeated across other data science companies and industries!

    edit: it’s in UTC and thus can’t really make conclusions based off of time of day. 🙁

  6. I’m not entirely sure what I expected to see. Perhaps a greater number of submissions before lunch, and a spike in the late afternoon for overnight computations. Maybe a Friday 3pm weekend computation.

  8. As a former DS, I can attest that Tuesdays are for most work, followed by Wednesdays

  9. They would do it Monday but the idiot bosses schedule a bunch of Monday morning and afternoon meets that fragment the day.

  10. If you make a heat map of work done by anyone in any field, it’s always going to be lighter on Fridays and Mondays because those days are much more likely to be holidays or vacation days.

  11. It kinda makes sense. On Monday you check your results that ran over the weekend. You adapt your code a bit, and ah shit its evening. You finish the next few details Tuesday morning, and then submit a new batch of computations to the cluster. It sits a bit in the queue, and then it starts computing.

    I definitely noticed this pattern, that when I submit jobs to the cluster they start faster on Monday and on Friday. In the middle of the week it usually takes a bit longer.

Comments are closed.