Tesla Software Engineer Interview Experience — Live NumPy Coding on 3D Plane Fitting

Second round for the sensing engineer position. The interviewer had me write everything in numpy the whole time. I hadn't prepared for that at all and couldn't really write it, and after 10 minutes I already didn't want to keep going with the interview, but the interviewer kept pulling me along and dragged it out for the full time anyway.

First question: given a list of 3D points and a plane represented as ax + by + c*z + d = 0, find the distance from each point to the plane. I didn't know the distance formula and I'd also forgotten the numpy API. The interviewer eventually told me the distance formula, and I stumbled through writing it out.

Second question: given a list of 3D points, 70% of the points lie on a plane and the other 30% are other objects or noise, find that plane and identify the points whose distance from the plane exceeds some threshold. I asked whether I could do an iterative approach — each time use all the points to run a linear regression for the plane's parameters, then remove the farthest 10% of points. The interviewer wouldn't say whether that was right or wrong, just told me to write it. I looked up some APIs, but no matter how I wrote the linear regression it wasn't coming out right. Then time ran out.

Felt like kind of a waste of time.