Imagine this. It's Friday night. Confidence is high. You're trying to hit the town with some of your friends... but how can you decide where to go? All of you have different preferences- after all, you don't know _ every _ restaurant in the city. There are so many options! That's where HMR comes in play.
A Happy Medium is Rare, or HMR, is a one-stop-shop social web application that takes preferences from a group of users and recommends restaurants. Just sign into Facebook, give us your preferences, and sit back to watch HMR work its magic! Ideal restaurants within a reasonable distance will be displayed beautifully for your viewing pleasure.
Behind the scenes of HMR are a finely-tuned series of blazing fast web technologies. MeteorJS was used to create the web app, and there was an effort to construct a recommender engine based on Yelp API data.
Parsing the huge amount of data provided by Yelp (not even analyzing it, just parsing it!) took a surprising and admittedly disheartening amount of effort. The data was so expansive that we could not hold it in memory and were forced to come up with workarounds. The real challenge however, I would say, was being tasked with building a recommendation system based on gigabytes of user data with 0 machine learning background. Every twenty minutes or so, we would realize there was another technology we needed to accomplish our goal, google it, flop it in the project, and move on.
Going into the project equipped with minimal knowledge on machine learning or recommendation systems or life in general made this a daunting task to say the least. Fortunately, Lucas came in clutch as a MeteorJS guru, and the rest of us were able to focus on the big-data back-end.
I'm proud of the entire team for not shying away from a challenging project. We knew it would be tough coming into it, but we embraced it as a learning experience, and went on with it anyway.