November 2024–present · Generative AI, agentic systems, RAG, Python and AWS
I own customer-facing AI engagements from discovery and requirements gathering through architecture, technical research, prototyping, implementation, deployment, demonstrations and stakeholder feedback. The work combines AI orchestration with backend engineering, interfaces, data pipelines and cloud operations.
¶ Government and defense
- SOCOM / adversarial AI: Designed, developed and demonstrated a platform modeling red-, blue- and grey-team interactions. Implemented agent coordination, supporting interfaces, environment preparation, VM conversion/deployment and MCP integrations with simulation tooling. The work included human-supervised adversarial and phishing/scam simulation concepts.
- U.S. Air Force RAG platform: Designed, built, deployed and demonstrated a chat and document-intelligence application supporting 100+ users. Used Docker and AWS GovCloud infrastructure designed around CUI handling requirements, with document ingestion, persistent knowledge stores, semantic retrieval and source-aware responses.
- Technical training: Helped design and teach a two-week applied AI/LLM bootcamp for 40+ Airmen, covering AI fundamentals, LLMs, agentic workflows and hands-on application development.
¶ Document intelligence and enterprise knowledge
- Financial documents: Built and demonstrated an application processing 1,000+ heterogeneous documents to extract financial information, aggregate amounts owed/paid and identify discrepancies. Designed side-by-side source review so users could verify AI-produced results.
- Invoice automation: Re-architected and repaired an existing ingestion application using Python, API Gateway, AWS Lambda, managed storage and validated structured outputs.
- RFP generation: Built ingestion and retrieval workflows combining customer RFPs with organizational knowledge to generate draft technical responses for human refinement.
- Research agents: Built file-system agents and document-question-answering systems for directory traversal, content discovery and semantic retrieval. Designed ingestion and knowledge organization for collections containing 700+ engineering documents.
¶ Agentic and multimodal applications
- Browser-observing agent: Built and demonstrated an application in which an LLM interprets browser/application state, provides contextual guidance and invokes tools for data extraction across frontend and backend services.
- Video-to-business-rules: Built a workflow combining recordings, frame analysis, transcription and LLM reasoning to produce structured, source-linked business rules for human review.
- Industrial modernization: Developed agentic workflows for Honeywell-to-Rockwell migration analysis using language and vision-language models, evaluating local and cloud deployment options.
Built Python and agentic tooling for COBOL-to-Java modernization. The pipelines turn code structure and relationships into graph representations and vector indexes for dependency analysis, discovery, semantic retrieval and modernization planning. Human-in-the-loop workflows support code conversion, Java generation, refactoring and technical review.
Deployed applications and demonstrations using AWS, AWS GovCloud, Docker, serverless services, REST APIs, Railway, Linux and CI/CD. Work included repeatable Open WebUI deployments and architectures intended to support hundreds of users.
As the organization's GitHub administrator, I support repositories, branching and pull-request workflows, automated builds, container images and deployment pipelines. I developed a shared LLM handler to standardize provider integrations and coordinated reusable functionality across repositories and submodules.
Internal automation work includes documentation generation, screenshot capture/processing, reusable Docker images and operational workflows.
As a research assistant, I developed an LSTM-based hearing-aid background-noise reduction pipeline and validated real-time inference on a Raspberry Pi 3B.
My master's thesis, AI Gait Restoration for Disabled Individuals, combined synchronized EMG, inertial and camera data with multimodal machine-learning experiments for gait prediction and restoration.
¶ Education and leadership
M.S. and B.S. in Computer Engineering, University of North Carolina at Charlotte; machine-learning focus/concentration and a Mathematics minor.
Volunteer firefighting and approximately two years of captain-level leadership experience, involving communication, coordination, training and operational discipline.
Profile · Projects · Homelab