KENNETH FINEGAN

An adaptable and results-driven professional who thrives both independently and in collaborative team environments, with a strong commitment to efficiency and goal attainment.

• Deed Price Calculator, Developer, Architected and deployed a public-facing web application using JavaScript, HTML5, and CSS to modernize the county's deed fee calculation process. Successfully migrated complex calculation logic from legacy 'nested' software into a standalone, responsive web module. Engineered a dynamic calculation engine that processes fixed regulatory values and variable user inputs to provide real-time, accurate pricing. This digital transformation significantly enhanced public accessibility and reduced the volume of in-person office visits by allowing residents to perform self-service calculations remotely.
• Vision System, Developer, Developed a computer vision system in MATLAB to automate target acquisition for a ping pong ball launcher. Engineered custom algorithms for pixel-to-metric conversion, enabling the system to identify targets, calculate real-time distance, and determine precise trajectory angles. Collaborated with a cross-functional team to integrate the vision module into a larger mechanical assembly, ensuring high-precision firing through coordinate mapping and geometric analysis.
• AIExpand, https://ai-expand.vercel.app/, Created and deployed AIExpand, a full-stack web application designed for AI-driven content generation and expansion. Built using a modern stack of Next.js, React, and TypeScript, the platform leverages Tailwind CSS for a responsive, high-performance user interface. Integrated advanced LLM APIs to facilitate real-time text processing and user assistance. Managed the full software development lifecycle (SDLC), including version control via GitHub and automated CI/CD deployment through Vercel, ensuring continuous uptime and seamless production updates.
• ImproveAI, Developer, Engineered a dual-LLM orchestration system designed to automate the stress-testing and refinement of AI assistants. Developed a synthetic data generation workflow where a primary model simulates diverse user inquiries based on business logic, which are then processed by a secondary assistant model. Integrated a Human-in-the-Loop (HITL) interface, allowing users to curate, edit, or void responses to ensure high-fidelity data. Built a persistent memory layer using a Vector Database to store validated interactions, leveraging Retrieval-Augmented Generation (RAG) to measurably improve the accuracy and context-awareness of future model outputs.