3 ms·
That's true, but engineers don't code whiteboards or codepens. If they want to know if I can code, I'm happy to point them to several git repos that I wrote >9
by dheera 7y ago
That's true, but engineers don't code whiteboards or codepens.
If they want to know if I can code, I'm happy to point them to several git repos that I wrote >95% of, and send code samples of non-public code that I wrote.
I can make real tasks happen, like make a robot navigate a room autonomously, but I suck at remembering how to write quicksort or write a parser for some weird interview-specific language or all of the insane number of C++20 features. At most whiteboard interviews at large companies, I was asked to code things I have never done for work and will probably never do for work.
Machine learning interviews frequently asked me to implement NMS or gradient descent. Fine, I can do that, and fumble a bit in the process (and probably get docked points for fumbling), but NO machine learning engineer on the planet needs to do these things because there are excellent libraries that do them for you and it isn't the job of the ML engineer to re-invent them.