• Designed and implemented a machine learning framework for Rate of Penetration (ROP) prediction, improving accuracy by 7 points (77% → 70%) over the existing XGBoost model.
• Conducted feature engineering and selection to reduce 52 to 21 key domain features, enhancing model interpretability and robustness across diverse drilling formations.
• Utilized state-of-the-art optimization techniques to conduct 60+ hyperparameter tuning trials, refining architecture depth, activation functions, and regularization to maximize model performance.
• Collaborated with domain experts to integrate machine learning insights into drilling optimization workflows, driving data-driven decision-making in operational planning.
