• Goal: To precisely predict the default rate of loan applicants based on the non-canonical background information they provided. • Role: I was the leader of the 6-person team • Result: Achieved Area Under ROC = 0.79780 (Private Score) with tree-based models. It was comparable to Kaggle silver medal's result. • Tech: Python (numpy & pandas), Hadoop-YARN-Spark cluster, Light-GBM modeling