Home

Systems with intent.

Systems practice

Workflow automation, internal tools, and bounded AI systems.

Building calmer systems for work that should feel clear, durable, and easy to run.

I design operational software, automation flows, and applied AI interfaces that reduce repetition, clarify handoffs, and turn fragile routines into maintainable systems.

View GitHub

Selected work

Work

CE-Assist

01 / 2026

email ops workflow feishu

An operational assistant for email analysis, state routing, notification logic, and bounded AI summaries.

XHS Factory

02 / 2026

content pipeline automation

A content pipeline for trend capture, topic synthesis, review, and publish-ready editorial output.

IC Database

03 / 2026

research data reference

A structured reference layer for semiconductor and component information, designed for lookup, comparison, and synthesis.

Open Source Contributions

04 / 2026

openclaw astrbot gradio

Hands-on contributions to community projects across AI tooling, agent systems, interface polish, and practical fixes.

Method

Approach

I approach software as an operational design problem. The goal is not to add more surface area, but to reduce noise, clarify decisions, and give each system a shape people can actually work with.

[01]

Operational Mapping

Start with the actual workflow: states, handoffs, failure points, and where people lose time.

Before building software, I map the operating path itself. Good systems remove repeated coordination, expose blocked states early, and keep the next useful action visible.

[02]

Interface Discipline

Design interfaces with a narrow purpose, strong hierarchy, and as little ceremony as possible.

I prefer surfaces that feel calm under pressure. The right tool should guide a team through work, not ask them to translate a dashboard back into reality.

[03]

Bounded AI Systems

Use models where they compress effort, but keep clear limits, review paths, and observable failure modes.

AI belongs inside guarded workflows, not loose promises. I use it where it sharpens execution, while preserving fallback paths, auditability, and human control.

The Lab

Ask

Single-turn interface

A guarded prompt surface for questions on systems, workflows, and applied AI.