Overall the interview went fairly smoothly. Four rounds total. The first round on stats was a bit shaky — if I were going to fail, that's where it'd happen. But it doesn't really matter much, I'd already accepted an offer from a good company elsewhere, so this one was purely for the experience.
Round 1 - Stats: They asked a few BQ questions, had me do a self-introduction, then gave an ML case — how would you build a model, starting from defining variables. I don't remember the exact question, but they went pretty deep, even asking me for the formula for the logistic function... felt like a lot for a product DS role. They also asked what "random" means in random forest, that one was fine.
Round 2 - XFN: with a head, about 30 minutes, overall felt pretty good. If you can handle Amazon's BQ questions, you'll be totally fine here — so just prep using Amazon's BQ style.
Round 3 - Case: ad-related, asked about the difference between adding a banner and not adding one — whether you need the banner at all. I defined some metrics like CTR, dwell time, banner click efficiency, and accidental clicks (had to define that one myself). Overall it went okay.
Round 4 - Case + Python: the Python part was just pandas, pretty simple — groupby, sum, count, filter, that kind of thing. The case was about the "group story" feature, whether to launch it. I looked at it in terms of potential usage efficiency, overall app usage time, and how often people post regular stories, and then they followed up with some A/B testing details.
Discussion
Loading comments…