HyperVerge Deep Learning Intern interview questions

First round was OA- where we were asked to train a classification model based on the dataset that they gave.. and check the accuracy of the model by uploading it to their API..(around 350 ppl were in this round) second round was GD- they asked if we were given a black box model (we dont know anything about it) and we can give it input and get output.. how will we "steal"/"replicate" the model.. (around 45 ppl were shortlisted for the GD) the third round was the interview this was very hard.. they asked concepts from the projects on my resume.. like explain CNN layer-wise, what is pooling in CNNs.. what is the formula for cnn, draw the diagram for CNN... differentiate between ANN CNN RNN.. things like this.. it was really hard for me as i wasnt prepared enough for this in-depth of an interview, that too offline..