it was a very toxic environment. The DS team existed for years within the clinical org, and then the CTO decided they needed to do more ML so created an MLE team. Originally it was pitched as an engineering team to create the pipelines to enable more ML by whomever (including DS), but the team members were more interested in solving organizational problems... but generally weren't equipped (connections, time, skills, whatever) to actually do that.
So, some of the "working on the same problems" was intentional-- it would require effort from both teams. But the dividing line was nebulous. The DS team would have preferred to do all of the stakeholder work through building a model, and then hand a pickled model to the ML team to implement in production. The ML team would have preferred to have the DS team scope the problem and hand off to them to do anything involving any form of modeling. It was a total mess.
But I have never worked at an organization where this has gone well, so I don't think it was an issue specific to that org. If you're involved in data things, you want to do interesting work and there's only so much interesting work to go around. And, ultimately, the vast majority of organizations don't have a need for tons of people to be doing the really technical aspects of ML/AI/etc. SO much of the work is scoping problems, cleaning data, worrying about pipelines, etc... and so if OP or whomever is thinking they're going to waltz into a job and make the next version of ChatGPT, that's really unlikely with anything less than a PhD. Personally, I've found a pretty good home being able to interact with leadership to define nebulous problems and solve those problems with whatever tool is appropriate-- and my success has way more to do with communication/project management/scoping skills than with technical skills (although both are necessary)... and I think those skills are better fostered through the more traditional programs.
Yeah that's pretty much my experience as well. I've also seen a lot of bait-and-switching going on where orgs tell candidates that they will be working on the cool interesting work, but then never deliver.
So, some of the "working on the same problems" was intentional-- it would require effort from both teams. But the dividing line was nebulous. The DS team would have preferred to do all of the stakeholder work through building a model, and then hand a pickled model to the ML team to implement in production. The ML team would have preferred to have the DS team scope the problem and hand off to them to do anything involving any form of modeling. It was a total mess.
But I have never worked at an organization where this has gone well, so I don't think it was an issue specific to that org. If you're involved in data things, you want to do interesting work and there's only so much interesting work to go around. And, ultimately, the vast majority of organizations don't have a need for tons of people to be doing the really technical aspects of ML/AI/etc. SO much of the work is scoping problems, cleaning data, worrying about pipelines, etc... and so if OP or whomever is thinking they're going to waltz into a job and make the next version of ChatGPT, that's really unlikely with anything less than a PhD. Personally, I've found a pretty good home being able to interact with leadership to define nebulous problems and solve those problems with whatever tool is appropriate-- and my success has way more to do with communication/project management/scoping skills than with technical skills (although both are necessary)... and I think those skills are better fostered through the more traditional programs.