Technical screening followed on on-site interviews. The on-site consisted of a presentation on an assigned topic followed by coding (I chose Python), design, and ML interviews. Overall I felt the primary goal was to gauge my clarity of communication and ability to respond to the interviewers’ questions. I interviewed during Covid, so being a clear communicator over Zoom is absolutely essential to success (both during and after the interview). It is better to show mastery of a few ML topics than attempt to be a Jack-of-all-trades — they aren’t going to push you on SOTA. The presentation was primarily a test of technical communication skill. I presented on an unfamiliar topic (auto ML), but I was glad to start the day on something focused on communication and preparation. I was asked a few questions at the end — none were tricky. The coding interviews were a typical CoderPad-style interview. Emphasis was placed on testing and communication. All the questions seemed to be leetcode easy/soft medium level. The design interview was focused on databases. I told the interviewer upfront that this is a weakness of mine, so we only discussed SQL databases. A more senior person probably would be pushed harder on this topic. The ML interview covered some basics (e.g. over/underfitting) as well as my previous experience. I described a previous project, including the successes, failures, and things I would have changed. Once again, emphasis on being clear, not on technical difficulty.