Onsite interview requires comprehensive implementations of traditional and recent ML techniques. The interviewees are not as willing to interact throughout the coding interviews. Received general ML questions on sampling and architecture implementation.
Very adhoc ML question asking to write code for that topic; the only way you can answer that question is if you have read the ML topic before.
The EM I met and the engineers all seem to be mostly fresh out of grad school with 2-3 years of experience. Not sure if I would trust the team if everyone is so early in their career.
Interview questions [1]
Question 1
Torch debug (very easy) and coding a ML technique (not the common ones but very focused on a deep learning method in a research paper) - it is easy if you know the topic
There is a take home test about domain knowledge (maybe), then is a coding interview (more machine learning/deep learning than algorithms). I stopped here so I don't know any more processes.
Interview questions [1]
Question 1
basic data processing such as evaluate model performance
