I'm a big fan of data in general. It can tell you a lot about what users are doing and help gain all sorts of insights. One such aspect is in making recommendations based on past history or others that have made similar choices. Years ago in fact I wrote a small app that I used among some friends to see if I could recommend wines based on how other ones were rated. It was a small app I shared among just a handful of friends, some with similar taste some with different taste. It at first was largely an academic exercise of writing a recommendation engine, but if I could find some new wines I liked along the way great. Turns out it was a lot more effective at recommending things than I expected, even with only a small handful of wines rated.
The other thing I'm a fan of is Postgres (not a big surprise there), and earlier today I was starting to wonder why couldn't I do more machine learning directly inside it. Yeah, there is madlib, but what if I wanted to write my own recommendation engine. So I set out