- An assembly question about a SIMD architecture that I cannot remember exactly. - Given an array of points (x and y coordinates), implement a data structure (a heap) for insertion and deletion of n nearest ones. - Implement a CUDA kernel that computes the frequencies of numbers given in an array - Given a very large file containing integers, implement an algorithm that outputs sorted version. (also do space and time complexity analysis) - Implement an optimized CUDA kernel for matrix transpose (also had discussions about GPU architecture) - Present a system you designed and implemented, and explain the changes for desired features asked on the conversation. - Various questions about deep learning like explanation of optimization algorithms, difference between gradient descent and stochastic gradient descent, underfitting, overfitting, gradient clipping and many more.
