TX-1 screenshot
TX-1Headless Enterprise OS
Active development — Phase 4A desktop-native agentic terminal for enterprise operations. Natural language in. Mathematically validated action proposals out. Built on Tauri, LangGraph, and Google OR-Tools.
Product Architect — AI Systems & Design
I design systems that act, not just inform.
Nine years building enterprise software taught me that the interface is rarely the problem. The problem is that enterprise tools tell you what’s broken and leave you to fix it alone.
I build agentic systems that close that loop — detecting failures, reasoning through solutions, and bringing the fix to the decision-maker in a single keystroke.
Enterprise software is at an inflection point. The shift from dashboards to autonomous agents isn’t a UI trend — it’s an architectural one. The products that will define the next decade of enterprise operations aren’t prettier interfaces. They’re systems that understand context, propose actions, and get out of the way.
That’s the space I work in.
Hero portrait or systems diagram
The work
Three operational prototypes. One underlying problem.
SS-1 screenshot
SS-1Signal Shell
Demo complete — active developmentA living market intelligence engine that monitors your competitive landscape continuously, simulates user reactions, and surfaces strategic briefs before your PM team has even seen the announcement.
Swarm-Lite screenshot
Swarm-LiteProject Studio
Demo complete — active developmentProduct teams gather user feedback late, expensively, and at the wrong moment. The structural cost of a user interview or focus group means the decision is already made by the time you have the data. Swarm-Lite removes that cost — answering the question in minutes, before the decision is made.
Writing
All articlesThe approval is the product
Human-in-the-loop is not a safety guardrail — it is a design decision with more product impact than any interface choice. The moment of approval is where the user evaluates machine reasoning and owns the outcome.
ReadThe notification-to-execution gap
Enterprise software is extraordinarily good at telling you something is wrong. It is almost entirely useless at doing anything about it. That gap is where the next generation of enterprise tooling lives.
ReadWhy personas beat averages
An averaged sentiment score of 3.8/5 looks like "broadly positive." But the underlying data might be two strongly positive, one conditional, one mixed, one not yet evaluated. The conditional and mixed responses contain the most valuable signal — and they are invisible in an aggregate.
ReadInterested in the work?
Each case study documents the full architecture — the decisions made, the trade-offs accepted, and what I would do differently.