WeRide Intern Data Scientist Interview Experience — OA, Three Rounds, and a Sour TL Round

I applied for the PhD intern position. First I did the OA, which was pure coding. After that HR called, and then they referred me to a team interview — the team I ended up interviewing for was the DS metric team.

There were three rounds total.

Round 1 was coding, three questions. The first was a SQL question — I forgot what it was. The second was pandas: computing a sliding window average, specifically: I have a pandas dataframe, and I need to compute the sliding window average with parameter k — for each row, compute the average of the previous k rows, the row itself, and the following k rows. For the first k rows and the last k rows, where the row doesn't have enough previous or following rows, set the value to -1. The third question was also pandas — essentially a LeetCode "merge interval" problem wrapped in a pandas + self-driving skin.

Round 2, the interviewer was based in China. We first chatted about my resume, then she asked a metric design question — something like, how do you measure the experience of riding in a self-driving car, what evaluation metrics could you come up with, that kind of thing, and she said a few more things about it. Then, based on one of the metrics I proposed, I said you could compare it against some map software's ETA versus the actual ride time used — she said to use that ratio. Then it turned out the ratio was very high, meaning the performance wasn't good, and she asked how I would investigate that.

Round 3 was the TL round. She asked about my resume, then some stats questions — how do you test whether two sets of data come from the same distribution, what sampling methods do I know — and finally an open-ended question: there's a lot of self-driving data from Beijing, and they built an evaluation system on it; now they're in Guangzhou and don't want to redo all of that, they only want to collect a small amount of data — how would I evaluate whether the self-driving system performs well in Guangzhou. Follow-up: if some scenarios in Guangzhou are very different from Beijing, what should be done differently when collecting the data.

I have to vent about this TL — she had this condescending, holier-than-thou attitude, and the whole interview experience was really bad. It felt like the recruiter had dragged her into it and she was just going through the motions, hard to really say for sure. I'd seen people mention before that WeRide's interview experience wasn't great, and now I've experienced it myself. The first two rounds' interviewers were both pretty good, but after this last one I finished the interview completely pissed off.