AI Systems Engineer
I’m a bench-scientist-turned-engineer in Isle of Wight County, Virginia. By day I keep a school district’s network alive; the rest of the time I build AI systems — voice agents, agentic platforms, and security audits — that have to survive an adversarial review before they ship.
// 01 — about
I started in a wet lab — two bachelor’s degrees in biology, chemistry, and biochemistry, and federally funded genetic-engineering research on cyanobacteria. Real science taught me one habit I never dropped: a claim isn’t true until the experiment that could falsify it has failed to.
That habit followed me into fifteen-plus years of IT — from rebuilding networks after storms to running district-wide infrastructure for public schools — and now into AI engineering as a partner at Bizaitec, an AI product studio, where I lead the automated security-audit practice and build agentic products end to end.
Everything on this page shares one design rule: no AI output is trusted until an independent model — and a human — has tried to break it.
"A correct 'no' is worth more than a comfortable 'yes'."
// 02 — method
My signature methodology — I wrote the paper on it. One model writes; a different model attacks the work; a human signs off. The writer of a change is never its verifier — and that rule is enforced in software, not just in principle.
"Test claims before asserting them. If you pitch a capability, run the control that could falsify it."
"Single-point outputs are suspect until verified."
"When in doubt: be honest, be specific, cite the source, and test the thing that could prove you wrong."
// 03 — constellation
My tooling turns codebases into 3D knowledge graphs. This is a sanitized constellation of the systems on this page — drag to rotate, hover a star to identify it. Rendered by hand on a <canvas>, no libraries.
// 04 — projects
Not concepts — running systems, each with a moment where a confident AI output failed a control and got fixed.
A multi-tenant AI phone receptionist, built solo end to end: live telephony, a retrieval brain that ingests a business’s website and documents, an admin dashboard, and an installable mobile PWA — backed by 200+ automated tests.
A personal AI operating system. Overnight, one AI agent proposes verifications of the day’s work, an independent model reviews them, and nothing is marked verified without human sign-off. It also audits my AI sessions for context rot and renders my knowledge as interactive 3D maps.
A Cloudflare-style entropy mixer you can use in your browser: CPU hardware RNG, your camera’s sensor noise, microphone noise, a quantum vacuum API, and the NIST beacon are hashed into one ChaCha20 seed. Picks Virginia Lottery numbers, checks them against official results, backtests “hot number” strategies (spoiler: none beat chance), and generates passwords and keys. Runs locally too.
Wrapped a 458M-parameter biomedical foundation model as a local tool server — then ran decoy controls that disproved its confident drug-binding predictions, keeping only what survived testing. AI as hypothesis generator, not oracle.
At Bizaitec I run the security-audit practice for our AI-built app suite — multi-round, read-only audits of authentication, authorization, and row-level security against live infrastructure, taking products from first draft to release-ready.
Fully local, zero-API-cost AI: a 128 GB unified-memory Ryzen AI Max+ running ROCm-accelerated inference on native Windows, plus a headless Linux GPU box (NVIDIA DGX Spark) serving multiple LLMs concurrently behind OpenAI-compatible endpoints — deployed, quantized, and administered by hand over SSH.
A 19-panel market dashboard where local models generate trade signals — until MAVL caught the LLM contradicting its own technical indicators. Fixed by computing the math server-side and letting the model only interpret. Paper trading; not financial advice.
Partner at Bizaitec, an AI product studio building a suite of business apps — including Flow Forge (marketing automation), HR Compass (HR management), and Recepta Desk (AI receptionist) — with security auditing baked into the pipeline.
// 05 — experience
Co-building an AI product studio: agentic business apps, an AI receptionist platform, and the automated security-audit practice that gates every release.
District-wide network modernization — fiber backbone, security segmentation, wireless redesign, and observability with proactive incident response.
Multi-campus infrastructure refreshes, automated device deployment, cloud migration pilots, and the monitoring that cut classroom-impacting incidents.
End-to-end technology for small businesses, non-profits, and ministries — secure networks, device management, automation, and long-term mentorship.
Federally funded cyanobacteria genetic engineering and field ant ecology. The lab is where "verify before you assert" became a reflex.
// 06 — arcade
Creative coding experiments in physics, AI opponents, and rendering — all in the browser, no install.
// 07 — contact
Network refresh, AI enablement, a product that needs a skeptical engineer — tell me about it.