I interviewed at Nvidia in December 2020.

The initial round was a phone screen with a manager. The screening round included general questions on computer vision, such as optical flow and its limitations, and vanishing point and its significance.

There was a coding question to check the merging point of two linked lists.

After this round, there was another round focusing on data structures and algorithms, and general C++ questions. This round included two problems from LeetCode.

Following the coding round, there was another round with a senior member focusing on computer vision. The discussion was regarding projects mentioned in my resume. Then, there were scenario-based questions, like estimating motion between two image frames and fitting a 3D plane given a set of 3D points.

I could not clear this round. I'm not sure what went wrong, as the interviewer also seemed convinced with the answers I gave and did not offer any counterstatements to my explanations.

1. Given an array and a number, find two numbers in the array whose sum is equal to the given number.

2. Given a perfect binary tree, connect all nodes at the same level.

3. Limitations of estimating motion using Homography or Essential matrix decomposition.

4. Limitations of the Kalman Filter in estimating motion between two frames.