By day I'm the lead cybersecurity engineer for a program office in the U.S. Space Force's Innovation & Prototyping Delta, where I oversee cyber engineers across several contractor teams and guide prototyping programs and space vehicles through RMF and ATO. Before that I spent six years on the GPS Directorate, where I helped achieve an ATO in a record five months that enabled a first-of-its-kind tactically responsive space mission — built and launched in under a year.
About two years ago I started doing AI evaluation and RLHF work on the side: ranking and critiquing model outputs across reasoning, writing, health, and open-ended tasks, and writing the rubric-based rationales that feed back into training. As a cybersecurity subject-matter expert I was pulled onto specialized projects grading models on offensive and defensive security tasks — penetration-testing workflows, secure code review, incident response — scoring each step for technical accuracy and safety.
That led to a contract on a frontier lab's safety & trust team, reviewing real-world attempts to misuse frontier models for cyber operations and adjudicating them against usage policy. My offensive-security background turned out to be exactly the lens that work needs — telling a legitimate, educational, or defensive request apart from genuine attacker intent, where both false positives and false negatives carry real risk.
That experience is what convinced me to pivot. I want to do this full time — help make models more capable, and help make sure they stay aligned, safe, and a benefit to humanity rather than a threat to it. This site is me building toward that in the open, from the math up. Read my full story →