• Spearheaded the end-to-end development of a real-time TV commercial detection system using PyTorch and Python, enabling local businesses to enhance customer engagement through targeted ad replacement strategies.• Collaborated with the front-end team to integrate scalable systems across diverse Android environments and TV equipment, optimizing and quantizing models for seamless deployment and rapid response times.• Conducted data analysis and replicated research to identify six key patterns and trends in TV commercials, strategically scaling datasets by 100x through data augmentation, which resulted in a 40% accuracy improvement in ad detection.• Streamlined SQLite database operations by designing a scalable processing pipeline to efficiently index, update, and match over 400 advertisement fingerprints, reducing system latency to 500ms and achieving a consistent 85% detection confidence.• Delivered daily technical insights and presentations to managers, refined strategies based on iterative feedback, fostered cross-functional collaboration, and ensured the timely completion of project objectives.

• Developed explainable large language models (LLMs) and machine learning models to analyze the mental health impacts of chronic Lyme disease, leveraging extensive social media data. Delivered AI-driven early detection tools for complex diseases, identifying over 10 novel symptom patterns through automated data processing, statistical analysis, and risk data analytics.• Led a portfolio of three concurrent machine learning projects, coordinating tasks among researchers and data scientists, managing timelines, and integrating feedback to ensure successful completion. Achieved a combined publication success rate of 90% in leading conferences, including the 2024 AI in Medicine and Machine Learning for Health Conference.• Designed and implemented a Lyme Disease and Mental Health Risk Classifier using Transformer models (GPT-based, BERT-based) and advanced binary classification algorithms. Achieved a 15% increase in classification accuracy over baseline tools, enhancing early diagnosis capabilities.• Maintained and processed a cohort of approximately 100K Reddit posts on Lyme disease by utilizing APIs, TensorFlow, and Python. Built automated pipelines for longitudinal data retrieval, enabling effective data-driven insights.

• Conducted reconciliation of invoices with purchase orders and receipts using Excel, ensuring 100% accuracy, proper documentation, and compliance with terms. Coordinated with procurement teams from over 20 corporate clients to resolve discrepancies.• Optimized financial data processing by entering over 1,000 invoices in one week and digitizing more than 60 boxes of financial records spanning three years using an ERP system. Ensured timely billing for month-end close and supported accurate revenue forecasting.