Engineered a reusable deep reinforcement learning framework with environment logic abstracted from model training (PyTorch). Optimized training throughput by 200% through multiprocessing and streamlined state management for faster experimentation and GPU utilization. Built a Flask REST API for live model inference, analytics dashboards, and embedded documentation, enabling non-technical users to interpret outputs and reducing onboarding time. Implemented GitLab CI/CD pipelines with Docker containerization to ensure consistent environments for model updates and web deployments.

Undergraduate Research Assistant at UVA School of Data Science

September 1, 2023 - Present

Led dataset construction and analysis for a university governance study presented at INSNA Sunbelt 2025 (Paris). Built a scalable ETL pipeline processing archival records from 700+ universities, producing a 1,000,000+ row dataset and reducing manual review time by 500+ hours. Automated document parsing with OpenCV-based layout detection and Tesseract OCR, achieving 95%+ extraction accuracy across 20,000+ documents. Raised data precision to 99%+ by implementing custom NLP modules for entity resolution, normalization, and error correction (OpenAI API, NLTK, spaCy). Analyzed leadership appointments with statistical modeling and network analysis (scikit-learn, statsmodels, NetworkX), identifying predictors of leadership selection and grant funding outcomes.