There are lots of fun problems in this-- our matching algorithm at the moment is not super complex. Once we have more feedback/data, we will constantly tweak, b

There are lots of fun problems in this-- our matching algorithm at the moment is not super complex. Once we have more feedback/data, we will constantly tweak, but the starting plan is to make more successful matchmakers weighed slightly higher over time. And obviously if by random chance two people recommend each other for each other, that'd get a pretty strong rating. We send matches to the users every week, so we're expecting we don't have to be spot on perfect each week. Some of the more interesting problems we solved was trying to make this widget as fast and scalable as possible. It's fed entirely from Amazon SQS queues and so has no database access what-so-ever. The entire architecture is designed to scale easily over multiple servers in stages as we need. Fun stuff all around. Thanks for the feedback!