Back to projects

Selected work

Selected public work in AI product development.

This page separates two kinds of work: a production multi-agent development system I created for Microsoft Copilot Mobile, and Atlas, my personal reliability lab for agent workflows.

The public evidence is intentionally bounded. Exact contributor and PR counts, Microsoft internal architecture, prompts, private code paths, review heuristics, and implementation details are not disclosed.

  • Production system In the first month, PM and Design teammates created and merged production PRs after engineering review; review capacity became the next constraint
  • Hands-on role Original builder, product owner, and active developer
  • Personal lab Atlas tests tool paths, evals, replay, and recovery
Abstract mobile AI development workflow board with phones, agent steps, and review paths.
Public adoption signal In the first month, PM and Design teammates created and merged production PRs after engineering review; review capacity became the next constraint.

Shipped production system / Microsoft Copilot Mobile

Multi-agent development system

I created and shipped a production multi-agent development system for Microsoft Copilot Mobile. It gives PMs, designers, and engineers a reviewed path from product intent to mobile changes: agents do the mechanical work, evals catch regressions, and review stays close to the diff.

  • Role Original builder, active developer, and product owner for the workflow
  • Signal In the first month, PM and Design teammates created and merged production PRs after engineering review; review capacity became the next constraint
  • Focus Agent workflows, eval gates, review capacity, source-of-truth checks, and recovery paths
Read project page
Abstract mobile AI development workflow board with phone screens, review paths, and quality gates.

Personal reliability lab / Atlas

A public proof package for a private AI system

Atlas is the private system I build and operate for real decision-support workflows. Its public page pairs a dated fact sheet with a worked eval case and simplified architecture, without exposing private records or operational details.

  • Role Creator, primary user, and evaluator
  • Shape CLI spine, agent layer, memory boundary, always-on jobs, eval harness, and replayable failures
  • Focus Tool-path discipline, observability, evals, and recovery paths
Open Atlas proof package
Abstract private AI workflow system map with a command-line center, eval loops, data boundary, and protected vault.