The interview was very easy. The task was to build a classifier with accuracy 80% with less than 1000 images using imagenet weights. I achieved an accuracy of 94% with very less images. They conduct interviews so that the HR can show counts and get their increments.