About
I work on agentic AI, mostly the security side of it. What happens when systems start acting on your behalf, how they hand work to each other, and why the trust model underneath most of it does not hold.
How I got here
I grew up moving. Honduras, Germany, Kwajalein, the United States, and more recently Portugal. Kwajalein is an atoll in the Marshall Islands, two square miles of land in the middle of the Pacific. It is where I learned to swim.
You get one thing out of a childhood like that, and it is not worldliness. It is speed. New country, new school, new language, new set of rules nobody is going to explain to you. Work it out. I picked up Portuguese in three months in Lisbon, and that is not talent, it is repetition. I have been dropped into unfamiliar situations my whole life and had to get good at them fast.
That is the only skill I have that transfers everywhere. I am not the guy who knows the most. I am the guy who is fine in a room he has no business being fine in, and who turns unknown into known faster than is reasonable. Everything else is downstream of that.
School went quickly. High school at 12, community college, William & Mary at 14, graduated at 16. Fine, but not the education that stuck.
The one that stuck was starting work at 15 and spending years around people two and three decades ahead of me. That is where I learned how decisions get made when there is money and risk on the table, why the obvious fix is rarely the one that ships, and how to make a case to someone who has heard a hundred worse versions of it. I taught programming first, then a startup, then several years in a regulated industry where I went from writing code to running the engineering.
I grew up rebuilding the house my parents bought, framing rooms and hanging drywall and dragging insulation through crawlspaces in July. None of that has anything to do with what I do now, and it has everything to do with how I do it. Once you have physically fixed something, you lose patience with solutions that only work on paper.
The research
I got to AI sideways. I was doing distributed systems security with Stephen Herwig, mostly protocol measurement, which is where my first paper came from. Then I took Antonio Mastropaolo's course and the connection was immediate. Everything I liked about protocol security applies when the endpoints are language models, except now the endpoints can be talked into things. That became my PhD work at William & Mary's AURA Lab.
I write here about agentic AI broadly. Delegation and authorization, since that is where the research sits, but also what it takes to run these systems in production, and the parts of the field I think are being measured wrong.
Publications
Search-Based Evolutionary Data Pruning for Class-Level Code Summarization. JAWS workshop at ICSE 2026. First author.
Resource-Efficient and Effective Code Summarization. FORGE 2025. Second author.
SIPS, IPPS, or Oops! An Analysis of the Security and Privacy of DNS Service Discovery. USENIX Security 2024, poster. First author.
Otherwise
I started tutoring at 15 and never stopped. Teaching runs through most of what I do, and it comes from the same place as the reason I like meeting people: I want to know what somebody knows that I do not, and how they came to know it. Ask a real question and then be quiet, and most people will tell you almost anything. That still surprises me.
It is why I travel alone. Nobody in the room has any file on you, and you have none on them, so the conversation starts from actual curiosity instead of from placement. I have gotten more out of an afternoon with a fisherman who quit at forty than out of most things I have read. Three languages makes that considerably easier.
I play rugby for a club here in Richmond. Three hours, twice a week, and it is the only part of my schedule I will not move. I lift most mornings. I have played guitar for years and I am good at it, which took a long time and a lot of bad playing to be able to say.
The rest of it is the Blue Ridge when the city gets small, and Lisbon, which I keep finding excuses to go back to.
Contact
Email and X. Always up for a conversation about agentic systems, and especially interested in hearing where you think I am wrong.