Leetcode style questions are not a bad backdrop as long as you do a few things. 1) pick questions that are actually somewhat aligned with a problem that would

Leetcode style questions are not a bad backdrop as long as you do a few things. 1) pick questions that are actually somewhat aligned with a problem that would come up for the role. Usually implementing a data structure of sorts is going to be way more predictive and relevant than a dynamic programming problem. 2) Ensure the question requires a fair amount and complexity of code to complete. 3) The question should just be a backdrop. Consider also how quickly and proficiently they can code. How intelligent they come across in conversation. Things they call out as side notes, testing, quality etc. Many interviewers seem to have forgotten the purpose of the interview is to be predictive to on the job success, not to invent some separate funnel and gauge how well the candidate did on that funnel. In practice, at scale, you will likely have enough correlation between success on a contrived interview system and general competency, but you're going to get a lot of false negatives/positives using that as a yard stick. My experience has been that leetcode "theory" is very weakly correlated with competency for most roles, and quality and speed of coding much more highly correlated. One of my best hires was a guy who couldn't implement a tree traversal in the interview