Honestly the "right" answer in most of these types of questions is to search for a popular existing implementation. There is never a "we've got to solve this bubble sort algorithm in a locked room with no internet access" scenario. I get that they want to select for people who can solve problems by thinking their way through solutions with code, but I've never been convinced that solving leet code challenges is the way to do that. I'd also much rather solve problems without writing a bunch of code that has to be maintained, so engineers familiar with the tooling ecosystem for their platform of choice should be highly valued.