◦ Drove improvements in cloud architecture, observability, and billing for OpenAI API requests, optimizing patient safety and operational efficiency.◦ As part of the Cognitive Computing team, responsible for integrating a NoSQL electronic health record with cloud infrastructure for training and scaling machine learning and generative AI in over 40 major hospital networks.

◦ With Product Semantics team, developed NLP models to customize product recommendations for customers, for complementary and substitute products.◦ Used React, S3, DynamoDB, CloudSearch, and Vis.js to develop a scaleable integrated dashboard for querying, displaying, and hand-traversing high-volume product graph data.

◦ Led the development and comparison of gradient-boosted machine learning models for peak usage prediction, advancing electrical grid measurement and verification tools for the U.S. Department of Energy.◦ Created a suite of data cleaning functions, analytics tools, and visualizations utilized by the whole research group at Berkeley Lab.

◦ As part of the Supply Chain Operations and Automation team, streamlined existing AWS data pipelines, added authentication features to factory devices, and improved computer vision architecture.◦ Developed a validation pipeline for component shipments from suppliers worth $50,000 annually.

◦ Scraped biomarkers to construct a patient cohort from electronic health records (100,000 patients).◦ Used scikit-learn and Stanford CoreNLP to develop a k-means clustering model to predict COVID-19 clinical outcomes from doctors' notes.◦ Designed and implemented an API pipeline via Kong to bridge new and legacy technologies at the firm.

◦ Collaboration with the Berkeley Institute of Data Science (BIDS).◦ Performed exploratory data analysis and natural language processing (NLP) techniques to analyze how state law affects firms from a marketing standpoint.◦ Built an advanced web scraper for the federal aviation accident database to construct a text corpus for NLP research.

◦ Built and deployed a data pipeline using Docker microservices to import PDF medical trial documents.◦ Extracted images, text, and metadata, processing over half a million pages of digital documents for deep learning applications.

◦ Performed time-series analysis on data from wearable sensors and 3D motion capture to differentiate between conventional and diseased respiratory activity. ◦ Used Google Cloud Speech API, Stanford CoreNLP, OpenCV, and TensorFlow to generate and process unstructured textual data in order to add sentiment-analysis to Electronic Health Records.