PORTFOLIOVOLUME 2026

I’m a designer who learned to build.

20+ years across brand, product design, and frontend. Today that is product UI, design systems, and the newest part of the job: how people work with AI and agents.

I take on work where craft and outcomes both matter.

CURRENTLY
  • role · UX/UI Design Team Lead @Supplyframe, a Siemens company
  • based · Los Angeles  ·  Hybrid  ·  Remote
  • pronouns · he/him
  • open to · Leading teams in product design, design systems, and design engineering
DESIGNING FOR AI
LIVE DEMO

Agents people can trust.

An agent is automation you set up once and let run on its own. Point it at a stream of incoming work and it flags the risky items, sorts and routes them, or follows up with someone who has gone quiet.

The hard part was never the automation. It was the screens people use to set it up. The people configuring these agents are everyday professionals, not engineers, and because an agent acts on its own, a setup screen they can’t read is a trust problem. If someone can’t tell what their agent will do, they won’t turn it on. That is the problem I design against.

INTERACTIVE DEMOLIVE
Pick a request. It comes back as a when / if / do instruction to approve before anything goes live. One of the four gets an honest refusal.
aiSet up your agentCOPILOT

Tell it what you want, in your own words:

Pick something above and watch it become a clear instruction you approve, before anything goes live.
Your agent will:
Nothing yet. Approve an instruction and it’ll show up here, in plain English, so you always know exactly what your agent does.

Six principles sit under it. None of them is about the model being clever. They are about the product being honest and legible.

  1. One word for one thing

    Everything the agent follows is an “instruction”, never a rule, task, step, or workflow. I locked the vocabulary across the whole feature, because switching terms is how people lose the thread on an already abstract idea.

  2. Nothing’s live until it’s accepted

    Every instruction, goal, and change arrives as a proposal with a button. The system says what it will do, never what it has done. On something that runs autonomously, that gap between proposed and live is the entire basis of control.

  3. The toolbox, not a checklist

    An agent’s capabilities are framed as a kit it was given, not a list of everything it wasn’t forbidden. A wall of pre-ticked checkboxes becomes a legible “here’s what this can do,” and granting a new power becomes deliberate.

  4. Honest refusal over silent failure

    When someone asks for something the product can’t do, it says so plainly and offers the nearest real thing. It never quietly maps the request onto the closest available tool.

  5. Choices you can picture

    When the agent offers options, they’re built from the user’s own tags, statuses, and fields: “tag it High Risk when the risk score is 80 or above,” not a vague “tagging.” You see the outcome before you commit to it.

  6. State you can read

    A configured agent describes itself back in plain language, and a review pass reveals dead weight, conflicting instructions, and gaps, so people catch problems they’d otherwise never notice.

BUILDS
05 PUBLIC REPOS

Code you can read.

  1. Enterprise Design System

    A hand-authored HTML/CSS system for dense product UI: 10 components, 3 compositions, light, dim, dark and high-contrast modes, WCAG AA. Its documentation runtime is compiled from strict TypeScript.

    Demo · See it live · GitHub

  2. Client Studio

    A self-hosted app for one-person service businesses (coaches, trainers, tutors) to sell session packages, get paid, let clients book their own times, send newsletters, and see how their website is doing, all from a phone. Booking is safe when two clients reach for the same slot, availability holds across daylight-saving days, and the newsletter builder's live preview runs the same renderer that sends.

    Next.js 16 · React 19 · TypeScript · SQLite · Demo · GitHub

  3. Analytics Dashboard

    Privacy-first web analytics you host yourself: a small tracking script and a login-protected dashboard. No cookies and no third-party scripts; a visitor ID is a keyed hash that changes every day, and the data lives in one SQLite file. Tracking a click takes no code, only one attribute on the link, button, or form.

    Next.js 16 · React 19 · TypeScript · SQLite · Demo · GitHub

  4. DesInspect

    A Chrome extension (Manifest V3) that puts a DevTools-style inspector in the side panel and saves every CSS and HTML edit, so it survives a refresh, a login, or a route change. I built it to work a design problem on a live site and come back to it.

    Demo · GitHub

  5. mtgunited-mcp

    The MCP server behind Magicians United, my full-stack side product, so AI agents can search and work with collection and deck data.

    Demo · Magicians United · GitHub

EVIDENCE
PEOPLE · SYSTEMS · CODE

What the work shows.

  1. FIG. 012003 TO NOW
    Independent2003
    In-house2013
    200320132026
    20+YEARS

    Design systems, product UI, and the brand work underneath. Independent practice since 2003; in-house and agency roles since 2013.

    Experience
  2. FIG. 02THREE TEAMS
    3TEAMS LED

    A studio of six at Dose Loui: two designers, two social media marketers, a photographer, and an SEO specialist. Three designers through fintech products at Cane Bay Partners. Two senior designers and a junior on the UX/UI team at Supplyframe, a Siemens company, since 2025, with interns every fall.

    How I lead
  3. FIG. 03LIVE, WITH CODE
    10COMPONENTS DOCUMENTED

    A design system of his own with live examples and copy-pasteable code for every primitive, plus three compositions.

    Design systems
¶ NOTE
VOLUME 2026
Updated whenever a piece is worth showing. No mission statement, no manifesto, just the work, dated and captioned, and the smallest amount of context required to understand it.
ALSOEarlier work in brand, illustration, and print.
Archive