The team is under iCloud and mainly works on time series and operations research optimization.
The first round was a resume deep dive with the hiring manager. They asked how to decide which cloud storage tier to use for a user's files, such as AWS S3 versus AWS S3 Glacier, since different tiers have different costs. It was essentially a classification problem: predicting those files' access probabilities.
The second round was standard ML forecasting questions, including time-series questions such as predicting daily demand for data center storage. How would you scale from one data center to 100? How would you handle the cold start for a newly established data center?
The most ridiculous part was that I interviewed in the morning, and just a few hours later that afternoon the recruiter emailed me an "unfortunately." I had never seen a rejection arrive that fast; previously, the fastest had taken one or two days. I felt the second-round conversation with the interviewer had gone reasonably well. I was a bit long-winded in places, leaving little time to ask my own questions at the end, and we ran slightly over. The round was only 45 minutes. But this wasn't LeetCode, and the interviewer seemed reasonably satisfied with most of my answers.
Discussion
Loading comments…