I totally agree with the sentiment in the article. Infact I wrote an article expressing similar sentiment "https://medium.com/datadriveninvestor/aint-no-thing-as-a-full-stack-data-scientist-e94a90db048a".

The availability of machine learning libraries and plug and play nature of these libraries have given the people an illusion that 'ML is easy' , it is merely importing some libraries and tweaking it till cost function goes down.

In reality it ain't so, the data scientist should really know what is going on under the hood of the algorithm, learn, explore and fiddle with the math.

The Devops while good to know for a Data Scientist, is actually a distraction from the real work.