Start at the failure mode.
The useful product is usually hiding inside what cannot be allowed to fail.
WORK / APPROACH / THINGS IN PROGRESS
I'm most useful where the problem is hard to name and the consequences are real.
My work sits between technical depth, product judgment, and operating reality. The recurring domains are AI, fraud, identity, cryptography, and security—but the method matters more than the category.
Full career record on LinkedInWORKING PRINCIPLES
Design for adaptation.
Fraud and security systems face opponents who learn. Static thinking expires.
Keep evidence close.
A decision should carry enough context to be inspected, challenged, and improved.
Stay near the implementation.
Architecture gets better when the people making decisions can still touch the system.
PUBLIC EVIDENCE
Ideas with a paper trail.
A small selection across invention and public technical work. Enough to follow the thread without turning this into a résumé.
GRANTED PATENT
Proactive Defense of Untrustworthy Machine Learning System
US 12,361,281 B2USENIX ENIGMA
In Security We Trust
Public talkBLACK HAT USA
Securing Your Big Data Environment
Public talkTHE BUILD RECORD
When a side project appears, it opens all the way.
No thumbnail-only showcase. Each build will include the product idea, technical details, source when public, and honest notes about the decisions behind it.
PROBLEM
The actual constraint and why it is worth solving.
PRODUCT
Who it helps and what changes when it works.
ARCHITECTURE
Components, boundaries, models, and trade-offs.
EVIDENCE
Source, demo, benchmarks, decisions, and current state.
NOTES
What failed, what changed, and what comes next.