After reading this article, it sure seems like they're counting people who are more on the research side of AI. And yet a lot of what I've been reading or hearing about has been almost just as focused on application of machine learning strategies, companies looking for people who are good at dealing with data sanitization and exploration, and then ultimately looking for people who can deal with creating the entire data pipeline needed to support applications that are powered by ML.
Because ultimately the "ML" portion of a large data-driven application is just one component. There's so much more to it ... is the data we have the right data? Where else might we find supplemental data to combine with our own? How will we keep our data fresh and cleansed? What data transformations are needed to stage it to be even ready to be consumed in a manner needed by certain ML models. What problem are we trying to solve? Will the data we intend to use actually help solve/answer that? ... those are the kinds of questions/problems that are incredibly important on any ML-driven app, not just which CNN or RNN model and how many layers are needed to support it.
Because ultimately the "ML" portion of a large data-driven application is just one component. There's so much more to it ... is the data we have the right data? Where else might we find supplemental data to combine with our own? How will we keep our data fresh and cleansed? What data transformations are needed to stage it to be even ready to be consumed in a manner needed by certain ML models. What problem are we trying to solve? Will the data we intend to use actually help solve/answer that? ... those are the kinds of questions/problems that are incredibly important on any ML-driven app, not just which CNN or RNN model and how many layers are needed to support it.
Wondering what others think on this topic.