We were standing in line to get food at the Hackathon when this idea struck us. We were thinking of ways to use Uber before, and this was the one way that seemed different than just making it easier to go from A to B.
What it does
A Whisk is an Uber that you have us specify, taking you to a place that we think you'll like. We use the same algorithm Netflix does, by classifying users based on their responses to a litmus test of places (we let people submit ratings for a wide variety of joints when they sign up), and then looking at how people in those categories rated places on Yelp. The goal is to expose you to new places and experiences wherever you live, and to make it easier for people to step outside their comfort zones.
How I built it
The frontend app was built in XCode, and made heavy use of the NSURL-related objects for the APIs. I used SciKit-Learn and NumPy to do the SVD, and we're hosting a Flask server to communicate with requests from the app.
Challenges I ran into
XCode is a lot messier with HTTP than we'd imagined (and with tables, too). Processing strings and JSON took some time in it as well.
Accomplishments that I'm proud of
The strength of the algorithm, learning Swift swiftly, and the portfolio of error messages we got over the course of the weekend.
What I learned
Swift, more machine learning theory, and more about hosting.
What's next for Whisk
Expanding to more locations, incorporating different kinds of data, and combining friends' preferences to let Whisk determine a hangout that's as fun as possible for everyone involved.