2nd round: This round was more into project and its related questions. know very well how the project works properly. End to end architecture was being asked. Asked me to build an architecture for a hypothetical project and how it works. Simple questions on python,pandas,nltk,numpy etc were being asked. They checked the basic concepts very very well here,and wanted to make sure that the person knows what he is doing. They expected me to know a little bit of angular JS, which i managed to tell something.

3rd round: -Fully basic level algorithms such as Regressions,Classifications(full),Clustering,Reinforcement learning, and a bit of ANN/CNN . End to end intuition was being checked here with cross questions and couple of mathematical equations. -Questions on statistical concepts, from very basic and easy ones to the unknown level: started with mean median mode to T-tests and degrees of freedom. A bit of variance,Standard deviation, random sampling distribution,z-score,Null value hypothesis,statistical significance were asked in basic, as i told I am not from stats background. - Also a bit of standard ML questions such as Bias-Variance Tradeoffs,False positives-negatives, ROC curve, AUROC,etc. -Full grilling in neural networks , more in CNN, to explain the stages in CNN. All the activation functions, and its differences and as to where and in what situations are they to be used. Full deep conceptual questions on gradient descent,stochastic GD,backpropogation algorithm,epochs, training a neural network,etc. -I cant remember all the questions here as each round lasted more than 90 minutes.

-Simple questions/code snippets on to select the highest value using only arithmetical operators,without using any conditional operators or any if/else/switch etc. -Questions on Travelling salesman problem and similar greedy optimisation techniques. -Simple Data structures questions of linked lists and trees. - Code snippet questions mainly picked from geeksforgeeks :D