Those who have faced ML engineering or applied scientist interviews so far, have you faced any traditional SWE system design questions ? Or, has it been more on ML system design side like "design the recommender system of amazon". While answering such questions, did you also touch base on the traditional system design concepts like load balancing, hosting etc. or, did you focus only on the Machine learning part and how to scale your models ? #machinelearning #machinelearningengineer #interview #systemdesign
Yes you should be ready for both general system design and more ML focused system design. I was asked to design distributed web crawler at a fang ML position interview
following
CVS interviews are a joke in general but Iβve been on the loop for our sad equivalent of ML engineers (basically looking for folks who have a vague concept of doing SVD on βlargeβ datasets). I tend to focus more on the design of the underlying system than particulars of algorithms/modeling strategies. YMMV at real tech companies.
What do you mean by design of the system? Just traditional SWE design concepts like caching load balancing etc. ?
Things like does the person prefer to move data or train in place, how data stores set up and why, for productionalization of model typical API stuff like load balancer etc., how are we bringing data into system, strategy for segmenting into test v training as appropriate, quantitatively define problem with reasoning, etc. So a bit of a mix but definitely system design elements for engineer more than for researcher