I design B2B products the way I always have: research first, accessibility non-negotiable, design systems as the source of truth. The AI era added range: I validate with real customers in Figma Make, then hand developers working demos built from the same Blazor components they ship to production.
01 · Skillset
Nine modules, each rated honestly. The meters fill to the level I can defend in a working session, not on a slide.
Edge design system component work, DTCG token specs with primitive and semantic tiers, and llms.txt files that teach the system's rules to the machines building with it.
WCAG 2.2 audits, ARIA patterns and focus-management specs for pickers, tabs and modals. It shows up unprompted in everything shipped, which is the only place it counts.
Prototypes that are real: single-file HTML systems on the design system's own components, with headless render QA in dark and light before anything ships.
Craft packaged as infrastructure: versioned agent skills for charts, decks and prototyping, plus agent instructions with governance built in from the start.
Precision briefs for machines and humans: scoped prompts that fix one bug and touch nothing else, behavior specs with state machines, explicit legacy-parity rules.
Survey instruments that survive critique, Nordic-language translation, five-session synthesis inside a week, and metric choices argued from context.
Building how the team works, not just what it ships: a vibe-coding working group with production gates, sprint facilitation and practice boards.
Designing for interfaces that talk back: MCP app components, tool-scoped assistants in team channels, and human-in-the-loop checkpoints drawn before the first wireframe.
GDPR-aware workflows, EU AI Act literacy, and a security review before any third-party plugin gets installed. Speed only counts if nobody's data pays for it.
02 · Track record
Built a responsible AI framework for invoice management and sketched the first agent concepts, before it was a job requirement. Governance became the backbone, not the afterthought.
Wrote llms.txt for the component library and a gated five-phase prototyping methodology, so AI output is design-system-faithful by default instead of by luck.
Survey landed, monochrome dashboard and branded readout deck shipped days later. Packaged the charting approach as a skill so the next study starts at the finish line.
A self-contained presentation engine: 15 animated templates, one brand token, verified contrast in both themes. Packaged as reusable skills and versioned in Git.
03 · Growth
SQL against real product data, cohort and funnel reads, and one instrumented feature with a defined success metric, a baseline, and a presented delta.
Git branching and pull-request review, enough JavaScript to spot a wrong pattern, and what CI actually checks. Directing AI is solved; reviewing what it writes is next.
Pricing, business cases and competitive analysis, learned on live product decisions rather than in the abstract.
Evals and regression checks for packaged skills and agents, so they are tested like software instead of trusted like folklore.
04 · Roles
A senior product designer focused on ethical, accessible digital experiences, uncovering unmet needs and the real drivers of value.
Senior product designer · Invoice Manager
Design lead for IMAdmin on the Edge design system. Owns accessibility across pickers, tabs and modals, contributes to the Blazor component library, runs usability research end to end, and drives the team's AI-assisted prototyping practice.
Role title · Team or product
One or two sentences on what you owned and what changed because you were there. Lead with the outcome, not the duties.
Role title · Team or product
Same pattern: scope, then impact. Older roles can be a single line; the skillset above carries the detail.
Next patch
Enjoyed this page?