• Formulated factory chemical reaction monitoring and optimization machine learning models (Random Forest, Decision Tree) from end to end. Using oxygen flow rate, pressure and temperature to forecast the reaction rate and catalyst usage. Developed interactive user interface using dash plotly to visualize live data monitoring and model simulation to help end users increase factory production efficiency. • Formulated soybean yield and quality prediction machine learning models (Random Forest, XGBoost and Linear Regression) from end to end. Using geospatial prediction in weather feature engineering and soil quality modeling. Designed hypothesis tests to identify leading indicators for yield prediction and seed quality prediction. Successfully built sellable seed forecasting model performing 17% better than historical average records. • Developed interactive user interface using matplotlib, seaborn, tableau and Dash Plotly to visualize data analysis, model predictions feature importance, and Shapley values. Provided strategic insights on seed growing, storage, and transportation to growers, contributing to increased efficiency and profitability.• Led model building activities and managed a team of 5 data scientists for the joint Soybean Analytics Project. Led road map development, implementing SAFe Agile methodologies, and Decision Science Life Cycle model maturity process to ensure ‘Quality Built In’ soybean yield and gemination forecasting models.

● Implemented big data solutions for distributed multi-thread CPU/GPU computing and ensure spectrums of data models running multiple platforms● Led data mining activities, including geological feature engineering and decision tree ensemble methods● Utilized machine learning tools, including non-parametric regression model and gradient descent method to develop and update 3D subsurface velocity model in supervised machine learning in 10+ pilot production projects● Developed new production flow with Full Waveform Inversion algorithm, analyzed the impact on velocity model updates, improved image resolution and geological consistency

● Applied prediction modeling approaches in data matching and surface related multiple attenuation in 3 production projects● Anomaly detection and supervised learning applied on seismic wave propagation, ruling out input outliers and assuring input data quality, improved signal-to-noise ratio● Performed seismic exploration data analysis on terabytes of data on company platform in 2 projects in Gulf of Mexico