Capital One Data Scientist Interview Experience — Four Random-Order Rounds and a CodeSignal Tech Round

Thanks to the earlier interview reports on this forum, I'm sharing the latest DS one here.

Interview process

Four rounds in random order: case / role play / job fit / tech, with a half-hour break between every two rounds.

Case

The classic burger problem. I vaguely remember a few points: a vegan burger sells for $5, so what are the overall concerns? Here you should ask what a normal burger sells for. I said customers won't pay the same price for a vegan burger as for a normal one, and then it was break-even calculations and so on.

Role Play

The classic airport problem.

Handling missing values, data formats, then choosing between linear regression and random forest. I picked linear regression and explained the coefficients, p-value, t-value, F-value, and how to improve the model.

Job Fit

Nothing much to say. Mostly we discussed what the team does, plus the classic camera question.

Tech round

This seems to have been changed recently. The old reports on this forum were all ML theory questions and past experience, but now it's CodeSignal with four questions.

Question 1: I was given a .sh file and asked what it does, and I had to run it. I could Google the parts I didn't know.

Question 2: a fit/transform class, and I had to explain what the class does.

Question 3: completing the test cases, writing just one was enough.

Question 4: debugging brand-new code. I didn't need to write code, just explain my thinking clearly. It was mainly about things like variable names and efficiency.

What happened afterwards

The HR ghosted me for a week, which felt like a bad sign. A week later the HR reached out: the original team is moving forward with other candidates, but I'm going into TM.