Build what I can explain
Test what I ship
Keep the useful parts sharp
Make the magic accountable
I work where product, engineering, and design keep interrupting each other. The goal is simple: interfaces that explain themselves, systems that hold up under real users, and AI features that stay useful after the demo.
The path so far has moved from privacy tools and file workflows into model gateways, agents, retrieval, and the practical plumbing that makes AI products behave in public.
I am not trying to make the work sound larger than it is. I am trying to make useful things, leave clean decisions behind, and avoid building software that needs a tour guide.
CLEX AI
Building one endpoint for model access, routing, keys, usage visibility, and playground testing, because every app eventually asks: which model today?
Privacy and utility tools
Recently shipped privacy and utility work across file movement, disposable mail, receipt parsing, and race-weekend intelligence. Strange mix, useful pattern.
Machine learning systems
Working through ML systems material while turning the theory into product infrastructure that survives contact with users.
Useful collaboration
AI builds, product engineering roles, prototypes, and projects where design and engineering need to sit at the same table.