Machine learning engineer

It was in morning around 11 pm -12 pm. Interviewer asked one of the project I worked on - the dataset details and metrics. After this discussion of one coding problem - Mirror Trees followed by discussion of semaphores/mutex and multi-threading in OS.

Machine Learning Basics, asked about k-means clustering limitations of k-means clustering, how these can be addressed. How to handle missing values in data, SVM linear kernel. What is the role of learning rate in training neural networks. How can you deal with overfitting problem. Logistic regression basics followed by it's MLE formulation. Difference between lasso and ridge regression. Difference between transfer learning and fine-tuning.

Machine Learning Coding, Started with asking convolutional layer and pooling layer in CNNs. Then interviewer asked me to write a code for convolutional filter in C++, first started with simplest convolutional filter (for 2D images) later he asked to extend the code for cases like padding and for 3D images, also asked to write a code for pooling layer. Followed by the questions related on pooling layer - advantages and disadvantages. Why only 3x3 kernel is mostly used in CNNs and not 5x5, also why kernel height and width size is same, can it be the case where height and width of kernel aren't same. Then asked to extend the code for kernel size (height and width not simiiar).

This was a Hiring Manager round. They told me about their team work and asked questions on that basis. Asked behavioral questions - have you faced a situation where you had to take the lead, tell me about a situation when you faced conflict with your teammates, how did you manage a situation where deadline was more important than quality if work.