I build the AI workflows a company runs on, and keep them working...

Sample projects: a shared company knowledge OS, an AI-run delivery pipeline, a support triage agent.
By background I'm a technical program manager: led delivery across 6+ teams and 20+ stakeholders
on business-critical programs tied to $600M+ in scope.

01 / Flagship projects

03
Flagship / 01

Org OS

A company's shared knowledge and workflows as one repo: every team has a folder, every repeatable task is a saved workflow, and anything that makes decisions gets tested.

For a company: 'Why did we decide this?' and 'what do we do when a customer site goes down?' stop living in one person's head. Anyone, including an AI agent, finds the answer in seconds, with the source cited.

  • Maps every team's work and marks each activity: automate, AI-assist, or keep human
  • Saved workflows for repeatable work: ticket triage, account research, meeting notes to action items, morning briefing
  • Decision-making workflows ship with test cases; one command returns an accuracy score and a list of misses
  • Rebuilds itself for any company: a setup command interviews you and generates the structure
Stack
Claude CodeCursor-compatibleMarkdown
  • Ships with a fictional 75-person industrial company inside, so the demo works out of the box
  • Answers must cite a company file or say 'no approved source, ask a human'
  • Includes a decision log entry where a test score changed the decision
Org OS walkthrough preview

02 / Supporting projects

04
Supporting / 01

Donekeep

An AI planner, live in production: speak or type, and it turns your words into tasks, routines, and calendar events.

  • Voice or text in; structured tasks, habits, and calendar actions out (LLM intent parsing)
  • Retrieval (RAG) over your own tasks and history, so suggestions come from your data
  • Nothing writes to your calendar without review; built and shipped solo
Stack
LovableLLM APIsGoogle Calendar
Proof