Implemented Retrieval-Augmented Generation (RAG) pipelines with LLMs to enhance decision-making in visa adjudication, mitigating threats of espionage by improving the accuracy and relevance of information retrieval.Leveraged Amazon SageMaker and Bedrock to implement and compare bi- and cross-encoder neural models, optimizing precision and relevance in Natural Language Processing (NLP) tasks involving complex queries.Optimized corpus data compilation used within the RAG pipeline, minimizing cloud resource consumption and reducing taxpayer-funded operational costs without compromising the relevance and accuracy of outputs.Containerized the RAG model with Docker and Amazon Elastic Container Registry (ECR), enabling scalable AWS deployment with auto-scaling policies for optimized resource utilization and consistent performance at scale.

Spearheaded the complete rebuild of a faulty PCB flashing station responsible for over $10M in product by using Python and hardware development to resolve a critical bottleneck and prevent over $50,000 in estimated late fees.Achieved savings of over $120,000 in project costs and 4,000 hours in required labor by optimizing the flashing process, reducing production time by 89%, and moving the estimated project completion date up by 14 months.Identified the need for and led the development of internal software to analyze assembly line downtime, resulting in widespread adoption by multiple engineering teams across North America to help eliminate production issues.Upgraded and maintained Python, LabVIEW, and SQL applications for 20+ product lines within an Agile team.

Developed a cost-effective, scalable LinkedIn profile search program using Python and the Google Custom Search API, reducing the firm’s reliance on third-party services and lowering monthly client outreach expenses by 35%.Improved efficiency by developing a program that leveraged Selenium to automate client lead engagement on LinkedIn, reducing the team’s time spent on client outreach by more than 4 hours per week.Enhanced the overall user experience and accessibility of the above programs by converting them from separate command-line interfaces to a single Tkinter graphical user interface, allowing for widespread use within the firm.Analyzed financial statements of local businesses to generate over 160 targeted 401(k) plan sponsor client leads.