Patrick Butler

* (Co‑PI) Evaluated various deep and classical prediction models for use in predicting “high acuity left without being seen” patients and presented findings which resulted in resource deployment to improve patient health outcomes (with Children’s National Hospital). * Led team designing Language Model (LLM) pipeline and Retrieval Augmented Generation (RAG) architecture using Langchain, ChromaDB, and ChatGPT to query open source documents for answering questions related to climate change; reducing manual labor by 99%. * Directed fellow researchers and graduate research assistants in a Natural Language Processing (NLP) and a information retrieval project to mine scientific arXiv papers for deep learning techniques: evaluated efficacy, and provided ongoing feedback throughout the research and project implementation phases. * Used open‑source data to predict real‑world population level events such as civil unrest and rare disease outbreaks in Latin America, making more than 30,000 predictions in 4 years that were independently validated as part of a $22M IARPA Project. Later, led the team that transitioned this project into a production environment in the government. * Coordinated VT team predicting and modeling COVID‑19 and influenza‑like illness outbreaks during pandemics using an ensemble of custom deep learning models, producing results which informs public policy as part of CDC’s FluSight program. * Researched novel methods of nuclear proliferation related event detection based on open‑source news data leveraging word embeddings to detect the semantic shift of words in the source material surrounding points of interest. * Taught seven semesters of CS5644, CS5664: 'Machine Learning & Big Data', and 'Social Media Analytics' – Covered basic machine learning techniques using scikit‑learn and big data related technologies such as Redis, Elasticsearch, and MapReduce.

* Ran successful consulting business, generating data‑driven results for clients using AI/ML, NLP, and signals processing techniques. * Designed and implemented algorithms to integrate sensors on nasogastric feeding tubes and used them to detect placement in the body using machine learning and signals processing techniques. * Designed algorithms to forecast the importance of new technologies based on semantic word usage using word embedding models. * Supported the EMBERS project at Virginia Tech and continued research for predicting civil unrest predictions and influenza like illnesses. * Advised startup on machine learning techniques to predict risk of re‑admittance in order to better place care navigators for high-risk patients.

Researching data mining, specifically temporal data mining with applications in neuro-science and bioinformatics.

Computer and Information Security