I don’t think it’s too much to expect data scientists to quickly learn some DevOps skills, as long as you can motivate the value for using them.
I don't just care about what you know now, I need to know your willingness and ability to think on your feet when confronted with a seemingly random puzzle and actually persevere towards an answer if necessary.
Not everything will fit neatly into the box of tools that were previously learned.
Not all smart people play chess, or know every line of Dr Who dialog, or are Makers, or like brain teasers.
Sounds like someone looked up a list of IT stuff you should know and applied it to data scientists randomly. In fact, most of those things in the list may or may not apply to a "software engineer".
TCP/IP: networking between cluster nodes Server setup: deploy a map-reduce cluster SOAP/REST: read/write data from services Software development life cycle: plan/deploy a reporting system for end users
He/she knows SOAP/REST but that unaware of that NN model.
A human can only retain so much. Invest in a team which has it's own specializations.
I don't think a data scientist need to know all that stuff to be good at his job
That said, I've run into interviewees who have simply refused to fully engage with a brain teaser -- like they just shut down and gave up, even after I provided hints. For me, that seems to be a signal that they're not likely to deal well with unexpected puzzles that appear in the normal course of engineering work.