I was managed to go directly to an onsite, presumably due to having previously done an onsite at Google and having many internal references.

The onsite was your typical Google technical interview, differing from the standard SWE onsite (which I had done previously) in that there were two ostensibly machine learning focused sessions.

The ML sessions were straightforward. I was asked a number of basic questions about deep learning (my specialty) and a somewhat open-ended system design problem. Apparently, I did not do very well on the design problem, which came as some surprise. I have no idea what they're looking for there.

Coding problems were par for the course.

The recruiter gave me more specific feedback than I was expecting, which only served to further my impression that Google's interview process is complete BS. I was told that one of my interviewers thought my knowledge of deep learning was 'superficial', which I thought was hilarious as their DL questions were themselves superficial and a quick look at my resume would have confirmed that my experience in DL is anything but superficial.

Another bit of feedback that I found particularly egregious: my last interviewer found my working linear time solution to be overly complex and not demonstrable of sufficient data structure knowledge. I don't really know what more one could want than a working linear time solution! And it really wasn't at all complex...

All in all, I would not recommend that anyone waste their time interviewing with Google. There are better companies with better interview processes. I have no intention of ever going through their ridiculous process again.

Good only for a free trip to California.