(1) To accurately interpret the data from drilling, developed fast elctromagnetic algorithms to simulate and process the data for oil drilling.(2) Built a ETL tool for the drilling data with Python. Implemented the algorithms above that ran parallel on CPU and GPU cluster to process the drilling data using Matlab, Fortran and C# .(3) With the help from the program I developed, an annual revenue increase of $43 million is expected by improving drilling efficiency and boosting oil production for oil company.

(1) Fracture Classifier- To reconstruct the digital core using the CT images, built a Fracture Classifier to indentify the fracture type on each image.- Collected CT images, performed data cleaning, preprocessing, feature engineering and selection with domain knowledge; trained and optimized multiple classification models including Random Forest, Unet, Trans-Unet, Res-Unet.- Built the Res-Unet model with a precision score of 90%. It doubles the precision of traditional methods and reduces computational time by 90%. With the help from Digital Core Reconstruction an annual revenue increase of $40.5 million is expected by boosting oil production.(2) Drilling Data Simulator- To simulate the data received by drilling sensors, built a Drilling Data Simulator with machine learning models.- Collected the data retrieved by drilling sensors in the field, performed data cleaning, preprocessing, feature engineering and selection with domain knowledge; trained and optimized multiple regression models including Logistic Regression, Random Forest, XGBoost and LSTM.- MSE of LSTM is minimum. Although the improvement in MSE is not very significant but it saved a large amount of time. With the help from this model, an annual revenue increase of $35.2 million is expected by optimizing the drilling parameters.