UCDOps® Agent
An AI agent that runs the entire user-centred design pipeline — transcripts in, working prototype out, with evidence traced at every step.
InteractionDevAnd the AI in between.
I design the interaction, write the production code, and build the AI layer that sits between them.
Selected work
From problem to production.
Nine live products — AI-first builds plus directories and marketplaces where structure and findability were the design problem. Open a case study or go straight to the site.
teslaway®
An independent AI wayfinder for Tesla — live news, model data, and ownership answers with sources attached.
UCD Chat
Free streaming AI chat for UX and service design. No signup, role-tailored answers, plus a UX interview simulator.
facechip
Upload a portrait, get three hand-drawn vector avatars in seconds. Processed in memory, never stored.
Tesla VIN decoder
A free Tesla VIN decoder — ISO 3779 validation, factory and drivetrain decode, recall check without ever sending the VIN.
UCD Coach
A daily design digest that reads the industry and returns the five things worth your time — sourced, dated, hype-checked.
UCD Surf
The searchable archive behind Coach — hundreds of briefings, full-text search, topic and date navigation.
DesignOps World
A hand-curated directory of 136 design systems from 42 governments and leading organisations — components, patterns, palettes, and Figma kits. Updated monthly. Free.
teslaway pre-owned
A pre-owned Tesla marketplace with transparent pricing, specs, and photos — available, reserved, and sold stock in one place.
How I work
Evidence-led research, interaction design, engineering, and deployment — practiced as one job.
Evidence before pixels.
Every product here started with a real problem and real source material, not a moodboard. When I build AI features, the output has to be traceable back to something a human actually said or published — that constraint shapes the architecture, not just the copy.
Design and code are one job.
I don't hand off. The interaction model, the design tokens, the component API, and the deployed front end are the same decision expressed four times. Doing them together is what makes shipping in days instead of quarters possible.
AI in the loop, human at the gate.
AI does the assembly. A person approves the result. Every agent I've built has explicit approval gates, because a system that can't be stopped is a system nobody trusts in production.
Ship it, then learn from it.
All seven products are live, indexed, and getting real traffic. Nothing here is a case-study mockup.
Let's build what's next
Scaling a design function, embedding AI into delivery, or taking something from zero to one — those are the conversations I want.
Designing the future. Operationalising experience. Scaling with AI.