PATRICK WALSH
RESUME/CV
Professional Experience
Lead Data Scientist
ECS Federal
Feb 2024 - Present
As a Lead Data Scientist working as a contractor with the Department of Homeland Security (DHS)'s Data Services Branch (DSB), I specialize in document classification and text analytics using advanced machine learning techniques. My work focuses on creating and training NLP models, including Naive Bayes, Random Forest, and Support Vector Machines (SVM), implemented through SciKit-Learn. These models are pivotal for classifying multi-page PDF documents, supporting critical DHS operations through accurate multilabel classification.
I also extract key entities from text using Named Entity Recognition (NER) and Regex, enhancing the precision and relevance of data analysis. Beyond model development, I deploy these solutions on AWS platforms like SageMaker and EC2, ensuring scalability and performance. Additionally, I perform UI development with Streamlit, enabling seamless document upload and automated analysis through a user-friendly interface, making actionable insights readily available across DHS.
Senior Machine Learning Engineer
The Cigna Group
Oct 2023 - Jan 2024
As a Senior Machine Learning Engineer at Cigna, I specialized in Text-to-SQL generation using Generative AI and OpenAI APIs. My responsibilities included formatting Teradata SQL queries, making API calls to Large-Language Models (LLM), fine-tuning LLM responses, and modifying SQL responses through procedural calls. Additionally, I served as an R&D AI engineer with the Research and Development team, contributing to innovative projects and exploring advancements in AI technologies. This role allowed me to develop a comprehensive skill set in text generation, API integration, and SQL manipulation, further enhancing my expertise in the field.
Data Scientist
ECS Federal
Jan 2023 - Oct 2023
As a contractor with the Multi-Channel Technologies division at the Department of Veterans Affairs (VA), I played a pivotal role in creating Generative AI applications. Notably, I developed a zero-shot model for automatically tagging case notes based on textual data. My involvement extended to both frontend and backend development of chatbots, utilizing Large-Language Models (LLM), semantic search, and vector databases. I employed FAISS, Chroma DB, and Pinecone for efficient vector database storage, utilizing them to process, chunk, and store documents as vectorized indexes, thereby enhancing the overall functionality and efficiency of the chatbots. I designed and implemented NLP models for analyzing and routing veteran
queries, employing various modern textual content classification techniques. The incorporation of Retrieval-Augmented Generation (RAG) significantly improved the performance of LLMs by leveraging indexed documents stored in a vector database, resulting in heightened user satisfaction and engagement. In addition to serving customers and business stakeholders, I utilized data to provide timely solutions to business problems. I extensively utilized Python for extracting, analyzing, and processing textual data, contributing to the development of deep learning multi-label classification models. Furthermore, I automated business processes using Power Apps, Power Automate, and Power BI, fostering increased efficiency and proactively identifying and preventing data discrepancies.
Data Scientist
SYSCOM, Inc.
Jul 2017 - Dec 2022
Data Analyst &
Arabic Linguist
U.S. Army
Feb 2009 – Mar 2017
As a Data Scientist at SYSCOM, I wrote Python code to conduct data wrangling and configure Deep Learning models, fine-tuning hyperparameters, and executing training jobs within both AWS and Azure environments. One significant achievement was the development and refinement of a computer vision (CV) model adept at accurately identifying and classifying nutrient deficiencies in plants. This breakthrough contributed to improved crop yields and reduced costs for farmers. I took charge of deploying and hosting this model on AWS, ensuring scalability and reliability to cater to clients worldwide.
Additionally, I played a pivotal role in the design and development of an innovative computer vision model that successfully classifies biofilm pathogens in microscope images, leading to a patented technology. My contributions extended to automating the data preprocessing pipeline for CV model development, significantly enhancing efficiency from a 2-hour process to just 30 seconds.
In the realm of Natural Language Processing (NLP), I created models for Topic Modeling, Keyword Extraction, Sentiment Analysis, and Named Entity Recognition (NER). My expertise also encompassed the development of CV models for image classification and object detection.
Engaging directly with customers, I interfaced to understand their business problems, build relationships, and provide effective solutions. In the realm of data management, I queried, designed, and updated relational databases using SQL and Python. Additionally, I developed graphing and visualization programs, utilizing Python, R, Tableau, and Power BI to analyze medium to large datasets. These efforts underscore my multifaceted contributions in the dynamic field of Data Science at SYSCOM.
In my role as a Data Analyst and Arabic Linguist at the United States Army, I demonstrated exceptional leadership by managing a team of 20 data analysts, contributing single-handedly to 25% of the product output—marking the highest volume for the entire division at that time. My responsibilities extended to translating Arabic text and audio into English for further analysis, delivering high-level classified verbal and written reports to customers. Additionally, I played a crucial role in decrypting secure digital communications to exploit organizational weaknesses.
In pursuit of operational efficiency, I developed and implemented streamlined data analysis methodologies, resulting in a notable 15% reduction in analysis time, all while maintaining the highest standards of data accuracy and quality. My leadership skills were further showcased as I led cross-functional teams in the successful completion of complex projects, ensuring on-time delivery of critical milestones and demonstrating adept project management skills. Recognizing the importance of data accessibility, I spearheaded the development of customized data visualization dashboards. This initiative significantly improved data accessibility and empowered the organization with data-driven decision-making capabilities.
Furthermore, I actively coordinated with government agencies, providing high-quality, time-sensitive technical support. My multifaceted contributions, from leadership and project management to linguistic and analytical skills, underscore my dedication to excellence in the dynamic field of Data Analysis and Arabic Linguistics within the United States Army.