Created a web crawler to obtain files with related experimental data, evaluating over 15,000 candidates and outputting 1,200 files.

Wrote Bash scripts on a Linux system to compile data for further analysis and flag areas of interest.

Created and tuned an XGBoost classifier to predict high-risk program categories with 93% validation accuracy.

Deployed model on AWS to aid customer outflow analysis and developed business recommendations supported by model findings.

Enabled the company to retain more customers by sending targeted discounts, saving over $10,000 in lost revenue.