- Improved accuracy from 45% to 85% for NER tasks by fine-tuning 3B and 13B parameter LLMs in the Python backend- Evaluated LLM architecture tradeoffs, thus improving per computation speed from 60s to 2-10s, while maintaining accuracy- Organized over 500 identified NER cases, outliers, and failure modes to standardize feature recognition in Java application- Oversaw CD/CI distributed computing, real time analytics, and model testing, while minimizing AWS severs costs by 50%

- Synthesized HIPAA-compliant 2D MRI slices from human 3D MRI scans by coding state-of-the-art Generative AI in Python- Utilized Computer Vision (CV) and supervised learning to efficiently track animal joints in Python with only 4 training videos

- Optimized and measured machine learning output data parsimony, cutting generative runtimes from 24 hours to 30 seconds- Wrote Local Field Potential algorithms in Python to simulate biological neurons based on 150 parameter combinations

- Shipped GUI and backend code to crawl 5-10 news hubs daily and retrieve industry news articles using the wx Python library- Doubled product visibility by shipping description-based inventory searching after staging a local WAMP sever for testing