You can find me on GitHub.
During the UCSD Data Mining Contest 2009 I implemented and tuned my own classifier. I experimented with the level of randomness of trees and large forests and ended up with a hybrid of a bunch of different approaches.
Since the results using my trees and linear Support Vector Machines performed really well, I took the project online.
As part of a Software Engineering-Methods lecture Mike Gist, Daniel Burrell, Dave Lawrence, Andrew Slater, and I developed a distributed whiteboard application.
The project became a lot more involved than we first thought, as we implemented a complete and branched history of every session and allowed all people to draw at the same time.
The project is still far from fun to use, but a lot of features are working very well. Have a look at the final report for details.