About

I'm an AI-systems architect with 32 years in enterprise IT — including 13 years at Accenture, where I reached Senior Director working with Fortune 20 clients — and a background as a U.S. Marine Corps veteran. Since 2020 I've run an independent AI consulting practice — applied-AI advisory and hands-on engineering for enterprise and SMB clients across a wide range of industries.

The work spans more ground than "AI architect" usually implies. I've built the security and governance guardrails that constrain what Copilot agents can touch inside a large Microsoft 365 environment — Purview, Entra ID, Conditional Access, the whole stack that turns responsible-AI policy into something actually enforced, not just written down. I've architected Copilot-powered virtual agents for a 5,000-agent public-sector contact center against a fixed legislative deadline, and delivered Teams Voice for 25,000+ users alongside it. I've built a production tool that queries eight independent public data sources in parallel and uses an LLM to synthesize them into a confidence-scored answer — the kind of problem where a single AI call isn't good enough, and the real engineering is in how you combine several imperfect signals into something a business can actually rely on. I've designed an autonomous operations agent that ingests live data from banking, CRM, remote-monitoring, and accounting systems and routes work without a human in the loop, and a privacy-first edge-AI appliance that runs real-time neural inference on a dedicated accelerator chip — no cloud dependency, signed event transport, and a full regression-test suite that validates the detection logic without needing the physical hardware.

What connects all of it is the same underlying discipline: don't trust a single AI output just because it sounds confident. Most people's mental model of "how good is AI, really" comes from a free chat app — that's the floor, not the ceiling. Real accuracy is a solved engineering problem: adversarial review (a second, independent pass whose job is to find fault with the first, not agree with it), grounding against curated source material instead of a model's memory, staged quality gates, and re-auditing content that's already live instead of treating correctness as a one-time check. It costs more and it's slower than trusting the first answer. That's exactly why most systems don't do it, and exactly why doing it properly is a real differentiator, not a talking point.

I'm currently working toward the PECB ISO/IEC 42001 AI Management Systems Lead Implementer certification — the standard covering how organizations actually govern and continuously monitor AI systems, not just deploy them.

I'm actively looking for senior AI leadership roles — architecture, engineering, or governance, at the Principal, Director, or VP level. Remote-first, open to hybrid; based in Oklahoma and staying here.

If you want to talk, reach me at [email protected], or book time directly below.