DC Daniel Chu

AI product leadership / hands-on systems

I lead AI product work and build the systems behind it.

I am a Partner Group Product Manager at Microsoft leading M365 Copilot Mobile product work. I also build multi-agent development loops, eval gates, review tools, workflow replay, and recovery paths.

This site documents the public parts of that work and lessons from Atlas, my personal reliability lab for agent tools and automation.

The through-line is making agent products dependable in real workflows, with evals, human review, and recovery built in.

View selected work See my product approach
  • Lead M365 Copilot Mobile product work
  • Build Agent development loops, evals, review, and recovery tooling
  • Test Tool use, memory, scheduled work, and replay in Atlas

Public workflow trace Simplified to show roles and review gates, not Microsoft internal architecture, prompts, code paths, or implementation details.

  1. Intent / contributor A PM, designer, or engineer scopes a product change.
  2. Agent development Agents help carry intent into implementation work.
  3. Eval Workflow checks make regressions observable.
  4. Engineering review People retain quality and product judgment.
  5. Recovery The path keeps traceability and repair visible.
A warm studio desk with AI workflow sketches, interface cards, and evaluation notes. Daniel Chu AI product leader building agent, eval, review, and recovery systems.
Public workflow sketch Agent development, product constraints, review paths, and repair in one frame.

Start here / Shipped system

Copilot Mobile Multi-Agent Development System

A public-safe view of a production development workflow that connects contributors, agents, evals, engineering review, and recovery.

View the project

Selected work

Selected systems.

Shipped system Agentic development

Copilot Mobile Multi-Agent Development System

A production multi-agent system that turns product intent into mobile code changes with evals, review, and recovery built into the path.

Personal lab AI reliability

Atlas

A public-safe proof package for my private AI system, with a dated fact sheet, a worked eval case, and a simplified architecture.

Contact

If you work on AI products where agents, evals, review, reliability, and recovery matter, I would be glad to compare notes.

Connect on LinkedIn See my product approach X