I applied online. The process took 2 months. I interviewed at Mujin (Tokyo) in Dec 2023
Interview
The first round was simple. Having a conversation with the recruiter and he was really friendly.
The second round was a technical round with the hiring manager and took about 4 hours. First, he asked about the fundamentals of geometric computer vision, wrote down the equations on paper, and showed them on the cam. Then ask about the pick-and-place pipeline. What camera will be chosen, and what technique to detect features and estimate the object pose? Also, ask about robot kinematics. Last, ask a leetcode-like problem (easy - medium).
No DL questions. They want to solve with traditional algos. Interviewers aren't punctual, coming in late. And this was for first stage. The recruiter seemed nice though. Good luck guys!
I applied online. The process took 4 weeks. I interviewed at Mujin
Interview
There were three rounds. HR -> Technical -> Meet with CTO. The HR round that was fairly straightforward. It was mostly about understanding salary expectations and if any visa sponsorship is required. The second round was a technical round with the Director of the Computer Vision Team. Here the questions were related to the fundamentals of Geometric Computer Vision. Questions related to Essential, Fundamental, and Homography matrices, Camera Intrinsics, etc. Importantly, there were no questions on Machine Learning based Computer Vision. While the Director clarified that they are happy to use ML based methods, I did not feel like it was a priority for them. They are similar to other traditional robotics integrators (FANUC, Cognex, etc) who are focused on factory automation.
Interview questions [1]
Question 1
What are fundamental, essential, and homography matrices and how are they different from each other? What are the components of a camera intrinsics matrix? How do you correct for lens distortion? What are the units of the elements in the camera intrinsics matrix? What happens to the camera intrinsics after you crop an image? How do you convert from 3D world coordinates to 2D mage coordinates?
Ask about the fundamental conceptual ideas of computer vision, like transpose matrix, and 3d reconstruction. And give a specific question, ask what is the process of getting the pose of the item.
Interview questions [1]
Question 1
given an item, design the pipeline to get the object's pose.
