How I shapeproducts.
My process connects user behavior, product structure, design systems, AI-assisted production, and QA, so ideas move faster without losing clarity.
Quality is inspected, not guessed.
Every product passes the same six checks before it ships.
Clarity
Every screen earns its space. If it doesn't help someone decide, it goes.
Accessibility
Designed with accessibility-first thinking, so the experience works beyond the average user.
Behavior
Tested against how people actually move through a product, not how we assume they will.
Performance
Measured, never guessed. Perceived speed is treated as part of the design.
Consistency
One system of decisions, applied the same way across every surface.
Validation
A/B testing, usability runs, and design QA turn opinions into measurable improvement.
Systems are shared decisions, not libraries.
A real design system carries decisions upward, from a single token to what the user feels. Versioned, documented, and owned like any other product.
Decisions travel upward: token to experience
Where AI fits in.
Figma, MCP, and local models move design into code fast. Every output still passes a human review gate before it ships.
Tracking what changes how products get built.
Product technology moves fast. These are the tools and practices I keep close, with product thinking at the center of all of them.
AI-assisted design
Speed on production, not on the decisions.
MCP workflows
Design context, piped straight into code.
Local AI / LM Studio
Private models. Nothing leaves the machine.
Design-to-code
Editable output, not throwaway generation.
Front-end practices
Clean, semantic, accessible by default.
Accessibility testing
Audited against WCAG AA, with AAA-level care where it counts.
A/B testing
Let real users settle the debate.
UX analytics
Watch where attention actually goes.
Motion as feedback
Movement that explains state, not decorates it.
Performance-aware UI
Perceived speed is a design material.
Design QA
The last 5% is where quality lives.
Reusable systems
Build once, ship everywhere.
One loop, run continuously.
01Understand
Learn the people, the problem, and the constraints.
02Structure
Shape the system and information before the screens.
03Design
Make it clear, accessible, and intentional.
04Build
Ship clean front-end, accelerated by AI where it helps.
05Test
Validate with real users and measurable signals.
06Improve
Feed what we learn back into the system.
Improve feeds straight back into Understand.