Uber Data Scientist Interview Experience — Rejected After a Tough BPS Round With AB Testing, an ML Case, and Coding

I have a master's degree and 1.5 YOE as a DS at a pretty chill, low-pressure company. Even though I got rejected, this was still one of the best interview experiences I've had.

HR Screening

They asked about my background on my resume, my visa/status situation, and my willingness to relocate. Then they asked why I was interested in the role, what tech stack I use day to day, etc. They also introduced the team background — this was for the mkt/ops AS team. They said they couldn't find anyone who could do similar modeling work and were hiring for it (my internal monologue: I'm not even close to qualified for this...).

After that they sent me a prep guide for the interview loop — BPS + TPS + six onsite rounds in total. Whoever can actually get through all of that is a beast.

The only downside was that I hadn't prepared for L4 at all, and HR didn't tell me what level I was interviewing for before scheduling the interview.

BPS

It was a one-hour interview with one of the interviewers — the first half was an AB testing + ML modeling case, the second half was coding.

The AB testing part was about designing an experiment for an email campaign — they asked about sample size selection, metric design, what to do if the experiment results aren't good, how to monitor the experiment mid-flight, etc. I hadn't prepared for this at all and my answer had zero logic to it; the interviewer was still really encouraging about it though, which made me want to cry.

Then they had me design an ML model to predict CTR — asked about imbalanced data, how to choose an evaluation metric, and what to do if the precision-recall curve and the ROC curve don't agree (I got this one wrong too, ugh).

Last were two easy coding questions — one SQL, one very simple stats-processing Python question. By that point my brain was fried from the earlier questions, I was slow to understand the problems, and my thinking-out-loud wasn't very fluent. I felt like I'd already failed by the time the interview ended. Right before rejecting me, the interviewer even said she enjoyed talking with me, which honestly just made me more embarrassed.

These days ChatGPT does most of my job for me, so my actual interview skills have completely atrophied. Aside from having decent business sense in the domain I work in day to day, everything else — case questions, coding, all of it — is worse than when I was a new grad. I do really want to work on marketing models though, so I guess I just need to practice a lot more.