In a data scientist interview, should I use Python or C++ for algorithm/data structure questions?

For a data scientist interview, you should be ready to write code in Python, especially if you list it on your resume. For algorithms and data structure questions the language often doesn't matter too much, but for data manipulation the overhead can turn what's supposed to be a quick 5-minute check that you know what you're doing into a half-hour slog that puts off the interviewer. Consider how much C++ you would have to write to do something like find the row with the largest sum in a tsv given the filename. I'll often ask a question like this in a data science interview just to check that the person can code. In Python, this is a fairly quick one-liner: max(open(fn), key=lambda row: sum(map(int, row.split()))) It can be a little longer if you're not as experienced with Python, but in Java or C++ it becomes a horrible mess of file I/O and string parsing and takes up a lot of the interview time. Very few people who use C++ in interviews are able to come out if this type of question looking positive. At the very least, I'll wonder why they would choose such a language to answer this type of question. For the more typical algorithms and data structure questions, go ahead and switch back to C++. Data scientists are always switching between scripting and more heavy duty languages, so this will either be ignored or seen as a positive. You can probably get through most of your interviews just sticking with C++.

What's the language most commonly used? Short answer: Python. Long answer: you're asking the wrong question. People running whiteboard coding interviews generally don't care what language you use, they want to understand your thought processes, not your knowledge of a particular language's syntax. So use the language that comes easiest, because that will focus the interview on the important stuff. If you really think the interviewer is going to care a lot about language syntax, then the machiavellian solution is to use the language with which the interviewer is least familiar. So, I would recommend C# for an interview at Google and Go for an interview at Microsoft...

I can't possibly imagine that any company would care that you did it in C++ vs Python.  In practice Python is probably better, but I think they would care more that you write clean code.    Usually it's easy to believe that someone who writes strong code in one language can write strong code in another.

Generally, you should use Python for data scientist interviews. I learned this the hard way. C++ is for creating high-quality, reusable code. Python is for getting stuff done fast (see caveat below). Unless you are as fast with C++ as you are with Python, you'd better stick with the latter. They really just want to see that you can answer the question quickly, not create elegant code. My ExperienceI was once given a take-home, timed programming assignment for a data scientist position and told I could use any language I wanted. When I told them I was going to use C++, they replied "Um.....well, okay. Are you sure you want to use that?" I guess that should have been my clue to make a different choice, but I figured I'd impress them with my software engineering knowledge. Dumb, arrogant mistake. CaveatI'm well aware that one can create high-quality, reusable code in Python and R (see The http://www.nostarch.com/artofr.htm for how to do software engineering with the latter).

Python is better for expressing ideas, ideal for discussions in a whiteboard. Less syntax clutter, more english-like language.   And it is a lingua franca among data scientists.   C++ and Java are more familiar to software  engineers,  and R is more familiar to statisticians.  But everyone can understand Python, to some degree at least.