01Machines
How do we know when a model deserves confidence?
I explore evaluation, competing explanations, and systems that can show their work—not merely sound certain.
Founder · Data & AI engineer · Builder
I build for problems that don't come with instructions.
Across machines, information, and human behavior, I'm drawn to uncertainty—incomplete signals, competing explanations, and ideas that have not yet become dependable systems.
I don't believe difficulty makes a problem important. But when a problem matters, its difficulty is a reason to begin—not a reason to look away.
Machines / structure in the noise
Information / meaning in context
Human behavior / intent in the gaps
Stanford Commencement · 2005
Steve JobsYou can't connect the dots looking forward.You can only connect them looking backwards.
Looking back, the through-line becomes visible.
- Machines
- Information
- Human behavior
- Dependable systems
One question, many forms
What deserves our confidence?
02Information
How do we find signal without hiding uncertainty?
I build ways to turn noisy, incomplete evidence into useful decisions while keeping ambiguity visible.
03Human behavior
How can technology support reflection without taking away agency?
My experiments in meditation, motivation, and human-centered agents ask technology to make room for people—not replace their judgment.
A through-line
The subjects changed. The question stayed.
Learning the signal
Start with what the evidence can actually say.
Machine learning, perception, APIs, and analysis taught me to look for structure without pretending the data is cleaner than it is.
Understanding the person
Build around attention, motivation, and choice.
Meditation, breathing, reflection, and multi-agent coaching experiments shifted the question inward: what helps a person see more clearly?
Making it dependable
Turn the experiment into something people can rely on.
AI evaluation, cloud delivery, privacy infrastructure, and open-source intelligence bring the same concern into production: earn trust with evidence.
Public work & writing
Proof you can inspect.
These are the parts I can show openly. The private work informs the questions—not the claims.
OSS Radar
Open-source ecosystem intelligence for package momentum and dependency risk.
Proteus
Self-owned private networking made practical, repeatable, and transparent.
Merge Theory
A research note on set-valued reasoning, consensus, and uncertainty.
Who checks if it works?