> I think the the same applies for many Data Scientist jobs, which are these days more about fetching/cleaning/visualizing data than making machine learning on

> I think the the same applies for many Data Scientist jobs, which are these days more about fetching/cleaning/visualizing data than making machine learning on it. Well, creating ML algorithms is a small part of data science. And getting data, and making sense of it, is a highly non-trivial part (though, somehow underappreciated). Very often it requires a lot of knowledge, skill and experience in used methods, dataset and statistics in general. Anyone can plug&play a scikit-learn algorithm within 2 lines of Python, yet it does not make everyone a data scientist. Anyone can copy&paste&run a deep learning network architecture, but without proper care it is likely to terribly overfit to underprepared dataset.