I applied for a ML position ( training and making for their systems ) for Perception.

Phone Interview:
Asked some basic Machine learning questions. Some interesting insights about thier system in tracking and detection.
Next I was asked what issues I faced in developing models and training them. (Trying to guage the practical experience)

Coding:
First asked me to do in C++, I said I prefer python for interviews.
Question was a common one from the list of lyft questions here. Asked to write in DFS instead of BFS for a popular BFS question. Which was simple but I hit a roadblock with time and the solution.
He responded to the thought process nicely. But was too focused in finishing sooner as he had some commitments (which is normal, no points here ).

Make sure to simulate yourself what might be asked in such interviews. Because if you are not prepared for some unexpected questions (non-technical, open ended) and are not able to describe everything that you thought you would say. You will not feel successful throughout. Just a general suggestion.

Seems like the C++ thing was a downer. Was expecting to code in c++, be sure to always make your interviewer happy. Anything can become a bad signal for them.