In times like the ones in which we currently live, making and maintaining our human connection is paramount to staying sane. And, much to humanity’s credit, there have been a ton of recently-developed apps and websites that allow us to do just that. However, none of the services currently out there allow you to evaluate the person on the other side of the screen based on both their personal software projects and their contributions to open-source.
Even before our current crisis, I was, frankly, shocked at how none of the well-established players in online matchmaking have implemented this much-needed feature. And, while there are some (so to speak) national heroes, such as myself, who include their github and devpost profiles in their Tinder or bumble profiles, we need to confront the unfortunate truth that, much to the detriment of society overall, this isn’t a common practice.
That is where Github dating comes in.
Thought I'd expand on this a bit to prove that the project isn't a complete joke:
To match users together, we first generate a feature vector for each user based off content from their github profile. This feature vector is comprised of some handpicked features (eg, which languages the user uses most) and some generated from a neural network trained on project READMEs (doc2vec).
We then save each user's feature vector. When it comes time to make matches, we run an approximate nearest-neighbors search to find the (theoretically) most similar users, and suggest those as potential matches.
Log in or sign up for Devpost to join the conversation.