Carbon Footprint Reduction - Multiclass classificationObjective: Reduce carbon footprint by predicting abnormal plant condition which triggers unprocessed shale oil leaksAction: Developed ML pipeline with pre-trained XGBoost to solve multiclass classification problem. Stacks: PCA, clustering, scikit-learn, XGBoost, MSSQL, MATLAB, Azure, statsmodels.Impact: F1 score of 0.75 with minimizing type II error by 75%. Actual implementation of the model reduced environmental impact by 12% annually, reducing a leak of ~40k barrels/year Production Optimizer - Clustering AnalysisObjective: Analyze the oil production based on noisy ~350 sensory data instruments and identify plant condition into low/medium/high performanceAction: Implement PCA enabled K-means clustering to segment the plant into 3 conditions and cross check the production efficiency on each clusters. Stacks: Jupyter, Scikit-Learn, Pandas, NumpyImpact: Created a new plant operational guidlines for on-site operation to take immediate corrective actions to counteract low performance condition. Uplifted production by 6.3%, $7.3M/yr