Scientist.
Builder.
Human.

I’m Vice President, Core AI – Data Science at Microsoft. Previously, I led data science for infrastructure at Facebook, working across the systems that support a global service. I care about the connection between a sound explanation and a decision that makes a measurable difference.

Earlier, at Stanford, I studied language: inference, uncertainty, and the factors that shape interpretation. That interest in what language supports connects to the questions I now ask of measurements, experiments, and AI systems: what is observed, what is inferred, and where does uncertainty remain?

An earlier chapter, in my own words: a career interview about infrastructure data science at Facebook.

I live in Seattle with my family. My interests include solar power, radio, outdoor adventures, and building games and stories with my kids.

DataInk is another thread: an exploration of turning data into carefully composed prints, influenced by Edward Tufte’s approach to information design. It asks what makes a visualization worth spending time with after the dashboard is closed.

This site gives those interests a shared home, alongside the professional work and a notebook that started years ago.

NLLabs is where I explore persistent AI agents, cognition, and systems with Clawd. The work includes experiments in continuity, memory, and trust boundaries. Selected observations appear here as Field Notes.

Aetla is the umbrella for exploring which experiments could become useful products. Its operating approach starts with the hardest uncertainty, limits the number of active builds, and asks what evidence would justify continuing—or stopping.

Informivore is one of those experiments. Across this work, I use AI collaborators for development and drafting; the account here distinguishes documented observations from ideas still being tested.

Language research, infrastructure data science, and agent experiments do not share a single scale or method. They do share a concern with the distance between evidence and a conclusion. I use this site to make those connections visible without flattening the differences.

The project map distinguishes a framework from a lab, a prototype from an available product, and a game from the tests used to build it. The reading notebook preserves earlier interests in their original context.

For professional conversations, please find me on LinkedIn. For new writing, subscribe to the RSS feed.

This is a personal site. The views and independent projects here are my own. I use AI collaborators for research, development, and editing; project notes identify the observations, fixtures, and open questions behind their claims.

Site and project status updated September 5, 2026.