Thanks for writing this post - I can relate to it so much. I've been in the industry for almost four years with a master degree in computer science. I spent my first 3 years working as the type of scientists who focused on fitting models and building dashboards for the business, and I thought that's the norm. However I was tired of being the middleman (having hard dependency on data engineer to collect the data/software engineer to deploy the solution) plus my value was mostly measured by how many models I shipped (and it took forever to ship due to limited capacity of SDEs) so I was burnt out and left. Somehow I ended up in a team, in which I became a full stack DS, but the team has no ML at all. It's engineering heavy and sometimes I even forgot that I'm a scientist (though I do take care of the analytical improvement part). I had doubts in the path forward but after reading your post, it looks like I somewhat was on the right track (per my definition). At the end of the day, applied data science is not about using the SOTA models; instead, it's to provide values through data in a consistent manner, hence engineering skills are a must. Time to tidy up my codes again :)
I've never worked in business-focused data science, but I previously worked in biological data science. At least in business data science, the future which you're predicting quickly
becomes the present. So there's some natural selection favoring truth and reality. Imagine a world where you're predicting 10 years (and two clinical trials) into the future. That is what computational biology looks like. Biological data science does get better the closer you are to clinical use (diagnostics and prognostics). But for biological hypothesis generation, it's hard to find anything that's not "making executives feel like they're living in the future" or "making data scientists feel like they're back in grad school."
I'm trying to incorporate some of this into my model of the relationship between different levels of corporate; when I was in retail pharmacy we were constantly being handed different arbitrary metrics and being told we had to hit *mumblemumble* target (or else we'll get *mumblemumble*'d)--is "data science" broadly where they're trying to get these metrics from? There's elements of this that feel familiar in the form of "upper level corporate seems to think they know what they want but they don't have a clear picture of what we're actually doing over here and so are completely incapable of asking the right questions".
Hmmmm yeah potentially. But it could also just be your normal run of the mill finance/business analysts who have an excel sheet that says if you hit X, Y and Z they will meet their financial goals.
I have Walgreens and CVS in mind here, my guess was always that they'd outsourced the numbers-conceptualizing in a similar way they were seemingly being sold some other mix of weird HR and corporate culture packages. I'd be surprised to find out they had an in-house department working on it but maybe less so if it's under the same sort of aloof-ness-of-direction we had on the ground floor.
Had a pharmacy manager at the latter who was very fond of using a sophisticated dashboard that could show me all the numbers proving I was Doing It Wrong but got stunlocked as soon as I asked her what specific things she wanted me to do to change the numbers.
From one “recovering data scientist” to another, thanks for writing this.
Thanks for writing this post - I can relate to it so much. I've been in the industry for almost four years with a master degree in computer science. I spent my first 3 years working as the type of scientists who focused on fitting models and building dashboards for the business, and I thought that's the norm. However I was tired of being the middleman (having hard dependency on data engineer to collect the data/software engineer to deploy the solution) plus my value was mostly measured by how many models I shipped (and it took forever to ship due to limited capacity of SDEs) so I was burnt out and left. Somehow I ended up in a team, in which I became a full stack DS, but the team has no ML at all. It's engineering heavy and sometimes I even forgot that I'm a scientist (though I do take care of the analytical improvement part). I had doubts in the path forward but after reading your post, it looks like I somewhat was on the right track (per my definition). At the end of the day, applied data science is not about using the SOTA models; instead, it's to provide values through data in a consistent manner, hence engineering skills are a must. Time to tidy up my codes again :)
Yeah, I quickly learned I never want to find myself in a position where I need to "rely on SDE resources"
I've never worked in business-focused data science, but I previously worked in biological data science. At least in business data science, the future which you're predicting quickly
becomes the present. So there's some natural selection favoring truth and reality. Imagine a world where you're predicting 10 years (and two clinical trials) into the future. That is what computational biology looks like. Biological data science does get better the closer you are to clinical use (diagnostics and prognostics). But for biological hypothesis generation, it's hard to find anything that's not "making executives feel like they're living in the future" or "making data scientists feel like they're back in grad school."
I'm trying to incorporate some of this into my model of the relationship between different levels of corporate; when I was in retail pharmacy we were constantly being handed different arbitrary metrics and being told we had to hit *mumblemumble* target (or else we'll get *mumblemumble*'d)--is "data science" broadly where they're trying to get these metrics from? There's elements of this that feel familiar in the form of "upper level corporate seems to think they know what they want but they don't have a clear picture of what we're actually doing over here and so are completely incapable of asking the right questions".
Hmmmm yeah potentially. But it could also just be your normal run of the mill finance/business analysts who have an excel sheet that says if you hit X, Y and Z they will meet their financial goals.
I have Walgreens and CVS in mind here, my guess was always that they'd outsourced the numbers-conceptualizing in a similar way they were seemingly being sold some other mix of weird HR and corporate culture packages. I'd be surprised to find out they had an in-house department working on it but maybe less so if it's under the same sort of aloof-ness-of-direction we had on the ground floor.
Had a pharmacy manager at the latter who was very fond of using a sophisticated dashboard that could show me all the numbers proving I was Doing It Wrong but got stunlocked as soon as I asked her what specific things she wanted me to do to change the numbers.
As someone just stepping my shoes into this world, I'm glad I read this now. Thanks for sharing.
good (and painful) read
positive business results (money saved and earned) are the best peer reviewers. Keep up the good work
Thanks for sharing. This is extremely relatable.