• Founded a novel LLM-based tutoring platform valued at $2,000,000 for high school and college students• Managed three Agile development teams focusing on UI/UX, backend, and LLM architecture • Increased course-related factual accuracy by up to 35% over standard LLMs by implementing a series of routed RAG-based knowledge banks, finetuned LLMs, and optimized routers. • Reduced sunk LLM API costs by 95% by optimizing performance on open-source HuggingFace LLMs with prompt engineering, finetuning, and hyperparameter optimization• Developed and architected a cache-optimized REST API RAG service using Fast API, PGVector, PostgresSQL, Redis, and Docker• Designed and implemented a React frontend using NextJS and integrated with backend services using REST API protocol

• Designed and implemented an ontology generation and validation framework to expand basic formal ontologies for specific domains, producing a domain ontology consisting of >50,000 elements • Trained synthetic datasets for problem decomposition using more than 2 million diverse ontology-based scenarios • Researched and developed automated intelligence sensemaking using RAG, Llamaindex, and Langchain pipelines and routers, contributing to the annual Internal Research and Development proposal• Developed a project-tracking feature for an advanced timekeeping application using Java Swing to ensure governmental compliance, increasing precision to a window of six minutes

· Developed and implemented a training pipeline for NLP models, utilizing hyperparameter optimization to achieve up to 20% higher accuracy and precision compared to existing models · Developed and implemented a full-stack web application with Figma and React, allowing for a tight integration that reduced overhead and costs by 25%· Designed and implemented a more time-efficient REST API for an NLP service platform, reducing latency by 60%· Wrote modular services for deployment to Docker and AWS, ensuring high availability, scalability, and data security· Analyzed data from language processing experiments and created projections to identify trends in model performance, leading to improved accuracy and precision of NLP models by up to 20%

· Utilized full-stack web application development to identify assets and integrate vulnerability reporting into a unified database, enabling significantly faster response times to security incidents· Leveraged virtualization and containerization to track malicious files in a controlled environment, utilizing IP geolocation and widespread tracking to detect and analyze unidentified connections to identify and analyze potential data breaches, increasing investigation speed by 50%· Automated the process of analyzing malicious email traffic, resulting in a 250% faster response time, 25% increase in accuracy, and a 30-hour reduction in manpower needed each week.