Shipped
Production AI systems running today — Claude in the decision loop, real users, real SMS, A2P 10DLC compliant. More live infrastructure than I had at a funded dot-com in 2000.
What I build →Security leadership → a safety layer for AI
I've spent 25 years making powerful, dangerous systems trustworthy enough to run the world's networks, banks, and governments.
First the streaming pipes and the DNS root. Then the banks and the governments. Now the AI layer that's about to sit on top of all of it. I'm not here to study this next one — I'm here to build.
- $500M+
- gov revenue closed at Gigamon
- L2–L8
- network visibility, down to the human layer
- 30+
- Claude sessions a week, in production
- 4
- patents in network visibility & streaming
The through-line
Build it. Secure it. Make it safe. Sell it.
Secured
Bloomberg's first CISO. Tapped the network at L2, L3, L4 — and L8, the human layer. Built correlation engines before SIEM was a category, and the back channel between the USG and Wall Street.
The record →Sold
Invented Gigamon's inline L2 SSL Intercept, drove $500M+ in closed government revenue, and rang the Nasdaq bell. I've been in the room where the business gets made.
Patents & deals →The pivot, in one paragraph
The model-layer work is extraordinary. The deployment-layer story is what's missing.
The builders of frontier AI have done remarkable work on model-layer safety. The part that's missing — and about to matter a great deal — is the deployment-layer story that lets a JPMorgan CISO sign off on putting this technology inside a regulated bank.
That story needs someone who has lived on both sides: the regulated-industry CISO chair and the actual code that integrates frontier models into production under compliance constraints. I've sat in their chair. I know what they're afraid of. And now I know this technology from the inside.
I don't talk about AI. I run it in production.