I spent 20+ years working in technical operations. Now I design AI workflows around a simple assumption: the model will get things wrong. In every system I ship, uncertainty goes to a human: when confidence drops, when sensitive data appears, or when the answer matters too much to guess, the system stops or flags it instead of guessing.
Vigil - Reads incoming IT support tickets, sorts them, and sends anything risky, sensitive, or uncertain to a human instead of guessing. Evaluated on 26 real tickets I labeled myself (a small first sample, not a benchmark): 88.5% right, all six tickets containing personal data caught, all three mistakes sent to a human.
Threshold - A resume tool that tells you no. In the app, a match threshold decides whether tailoring runs at all. Grounding checks flag anything the source resume does not support, and you decide what to keep. Export produces a clean 2-page PDF or Word doc.
Most AI demos show what the model can do. I build for what happens when it shouldn't be trusted. The AI proposes. Rules and humans dispose.
- 20+ years in technical operations, service delivery, infrastructure support, and operational leadership, at CIMSystem, Accucode, CompuCom, and The Newcom Group. Now focused on AI governance, workflow automation, and human-in-the-loop systems.
- ISO/IEC 42001:2023 AI Management System Fundamentals, plus applied AI certifications
- Louisville, KY, working remote


