At Hoover Rehabilitation Services, I designed and deployed a config-driven Python framework that automated daily SFTP fetching and decryption. It now runs six scheduled jobs across three external connections, with new jobs onboarded through a JSON configuration file.
I also reduced a daily document process to one Python script that converts about 1,800 TIFFs to PDFs, and corrected a recurring data load that had grown a lookup table to 370 million rows. I improved stored procedure performance across eight procedures, cutting runtime from about 15 seconds to under 0.3 seconds.
In my personal Game Trailer Engagement Pipeline project, I modeled about 85,000 YouTube comments so changing engagement metrics can be refreshed without rewriting immutable comment text. For my Text Classification Research project, I built a Naive Bayes classifier that achieved over 95% accuracy distinguishing Charles Dickens text from GPT-3-generated text.
