Projects
3 key projects demonstrating technical depth and problem-solving capabilities
Problem
A traditional résumé can't demonstrate the one thing it claims — the ability to ship AI systems that reach production. Recruiters and hiring managers can't interrogate a candidacy on their own terms, and "I build with AI" is an unverifiable claim without a live artifact behind it.
Solution
Designed and shipped — primarily with AI coding agents under a spec-driven workflow — a production AI web application that is itself the portfolio: RAG chat grounded in real résumé data, a real-time voice agent, a Model Context Protocol server, a job-description fit analyzer, and a graduated prompt-injection defense, engineered like production software (split FastAPI/Next.js stack in Docker, local CI, and a parallel plan-driven execution workflow).
Outcome
A live, server-rendered, fully-indexable site that is the proof itself — AI-assisted delivery that reaches production rather than a stalled pilot. Recruiters, hiring managers, and prospective clients can explore the background conversationally or by voice, paste a job description for instant fit analysis, or connect their own AI agent over MCP.
Problem
At a designated Qualified Health Information Network (QHIN), audit logging was embedded inside a monolithic nationwide health-data-exchange platform and wrote an ever-growing volume of audit records — roughly 20 TB — into MongoDB. Storage cost climbed continuously, and a mandatory, retention-sensitive compliance concern was tightly coupled to the core application.
Solution
Redesigned the audit-logging subsystem into a standalone service: built a proof of concept, ran performance testing, and integrated an external storage vendor for durable off-site retention. Migrated approximately 20 TB of audit data out of MongoDB, decoupling audit logging from the core platform so it could scale and be retained independently.
Outcome
Moved roughly 20 TB off MongoDB and off-site for significant storage cost savings, decoupled a business-critical compliance concern from the monolith, and preserved the regulatory auditability a QHIN requires.
Problem
In an FDA-regulated cardiac telemonitoring operation, clinical technicians picked up monitoring review work in an ad hoc way. Distribution was uneven and the manual step was a throughput bottleneck for business-critical clinical software, where review turnaround directly affects care.
Solution
As business analyst, technical lead, and architect, designed and implemented a work-assignment component that systematically routed and assigned clinical review tasks to technicians — balancing the queue and applying prioritization — and bridged clinical stakeholders and the engineering team to get the requirements right.
Outcome
Boosted technician productivity by 85%. The contribution across business analysis, technical leadership, and architecture was recognized with a Certificate of Appreciation.