Kaggle competition in which one had to build a machine learning model with extremely less data that doesn't overfit on the testing data which was 100 times bigger than the training data. Final ROC score: 0.858 Final rank: 504(top 22%)