From my experience I kind of disagree that you can weed out non-programmers by looking at their Github profiles. This might work for people that have a lot of p

From my experience I kind of disagree that you can weed out non-programmers by looking at their Github profiles. This might work for people that have a lot of projects/contributions there, but it doesn't work for most Github users which have (if at all) only a few stargazers and no very active projects at all. So if we'd use Github data to weed out people based on their repos/stargazers count we would probably miss a lot of very good programmers that just don't happen to be active there. A simple coding test (ideally in a live environment over Skype) on the other hand allows me to (very roughly) judge a programmer's general ability within 10-15 minutes: Does he/she know the standard libraries of the given programming language? How does he/she structure the code? Simple problems that are roughly related to your field of work are suited best for this from my experience. For example, as we work a lot with graph data, one task we often use is to load a set of lists containing vertices and edges from a JSON file and convert that into a linked graph structure within Python. When hiring senior people most of this is usually not applicable as you can rely more on recommendations and the previous work experience of the candidate, which should give you better information than Github profiles / live coding tests.