Consumer Loan Product (Ant Jiebei) Cash Flow Forecast - Built Time Series forecasting model with Python and Prophet for products with a loan size of more than $100 billion. - Conducted time series data analysis, transformation and feature engineering with Python, Pandas and Plotly.- Implemented a customizable walk forward cross-validation API in Python for time series modeling in Scikit-learn.- Used the model in auto-loan servicing of more than 30 million users, increased the accuracy from 70% to 90%.Payment Fraud Detection System - Built data pipeline with Numpy, Pandas and Matplotlib for data preprocessing and EDA.- Selected features using PCA and visualization to identify patterns of fraud payments.- Implemented and tuned Random Forest, XGBoost and LightGBM models, final performance ranked top 5%.

- Conducted data analytics to improve customer segmentation/precision marketing models for 50M+ customers.- Optimized data streamlining process to automate repetitive data mining procedures, saved 1 FTE.- Used 40M+ data to build customer purchase prediction model that helped increase CTR for SMS marketing by 20%.- Developed a ML pipeline with Scala, Spark and Hive for data wrangling, feature engineering and model training.