Twenty years building enterprise systems for the Department of Defense and Fortune 500 companies. Now applying that same engineering discipline to AI-native applications that solve problems with real human stakes — starting with Medicare.
Two decades of shipping production systems in environments where failure has consequences — defense logistics, pharmaceutical sales, and Army-scale data migrations. That experience informs how AI Patient Advocate is built.
Solutions Architect and Lead Developer on the Army Logistics Modernization Program (LMP) — one of the largest SAP implementations in U.S. government history. Built dozens of Oracle APEX web databases that became the operational backbone of the program. Led a 14-person development team. Spearheaded multiple large-scale data center migrations from on-premises to cloud. Most recently: SAST/Cybersecurity Lead using Fortify on active Army systems.
Lead Developer and Technical Lead on projects for Merck, AstraZeneca, and Lockheed Martin. Designed and built physician-facing drug sales tools, project management systems, and repair-workflow platforms. Ported legacy systems to web-based architectures (ASP, Oracle, Crystal Reports). Early career: sales and marketing with Lincoln National Corporation.
— established proficiency
— AI engineering (2024 — present)
"As a caregiver to a disabled Medicare beneficiary, we spend an enormous amount of time reviewing and adjudicating claims from many doctor visits. I built this to change that."
AI Patient Advocate was not built as a portfolio exercise. It was built to solve a specific, daily problem: a disabled Medicare beneficiary generating dozens of Explanation of Benefit records per year, each requiring manual review to catch billing errors, miscoded procedures, and wrongful denials.
The administrative burden of Medicare is enormous and falls hardest on the people least able to bear it. A beneficiary shouldn't need to understand CARC codes, 42 CFR citations, or the FHIR adjudication schema to protect themselves from a billing error. That's what the tool does.
The application is built to production standards from day one — because real people will use it, and accuracy in healthcare data is not negotiable. Every number on screen is traceable to its exact source field in the CMS data. Every calculation is verified against the raw API response. A CMS production access application has been submitted.
Built in parallel with AI Patient Advocate — the same engineering discipline, different problem domains. Each project uses AI to remove friction from a high-value manual process.
Complete B2B pipeline that finds defense contractors from SAM.gov award notices, generates personalized opportunity digests, sends cold outreach, and processes Stripe subscriptions at $199/month. Includes an autonomous AI CEO (Victoria) that runs evaluation cycles and makes go-live decisions independently.
LangGraph agent that monitors SAM.gov for contract opportunities, scores them against target profiles, crawls procurement data across multiple sources, and generates personalized daily lead reports.
Cloudflare Worker that scrapes state procurement data and generates personalized cold outreach emails for contractors. Deployed and running.
CMS Blue Button 2.0 integration, FHIR R4 adjudication engine across all 8 Medicare claim types, 291-code CARC database, appeal letter generation, local LLM explanations. Production access application submitted to CMS.
Sophisticated medical RAG system with intelligent query routing, patient record anonymization, timeline analysis, and connection-finding across complex medical histories using vector embeddings.
Parses DISA STIG XCCDF files, maintains per-application documentation stores via vector search, and uses Claude to automatically assess compliance by combining STIG requirements with application-specific documentation. Built for DoD contractors.
Ingests documents from an inbox folder, chunks them hierarchically, stores in ChromaDB, and provides semantic search and Q&A via CLI and web interface using local Ollama models. No cloud required.
Automates STIG compliance assessments by reading requirements from XCCDF files and generating structured assessment reports for DoD system owners and contractors.
Takes photos of items, uses Claude Vision to identify them, researches comparable sold prices on eBay, calculates profit margins across platforms, generates platform-specific listings, and auto-posts to eBay via API. Complete photo-to-listing pipeline.
Chrome extension that summarizes web pages with customizable personality styles and community-specific perspectives (healthcare, faith, education). Cloudflare Worker backend with structured output.
Dual-profile relocation tool using Claude to discover and rank US towns based on user-defined criteria — cost of living, healthcare access, climate, community fit. Working Streamlit interface.
Built across government contracting, healthcare, compliance, and commerce — each one applying AI to remove friction from a high-value manual process. The common thread: production-grade engineering discipline, not demos.