Process

How I think and work.

A repeatable way of turning ambiguity into usable experiences — the same method whether the medium is a product flow, a learning experience, or an AI-assisted prototype.

FramePromptDraftEvaluateRefineHumanjudgmentAI accelerates the loop; a person decides what ships
My AI-assisted loop — acceleration with a human quality gate

How I approach ambiguity

I treat a vague brief as the first artifact to design. Before proposing anything, I write down the actual goal, the users, the constraints, and what success looks like — then I confirm I’m solving the right problem, not just the requested one.

How I frame problems

I separate the problem from the feature. Most requests arrive as a solution; I trace them back to the underlying user need and reframe them as a design question. Good framing makes the rest of the work faster and the decisions defensible.

How I use prompting across research, design & build

Prompting is a design medium, not a shortcut. I use it to widen the option space quickly — synthesizing research, drafting interaction copy, exploring alternative flows, generating variants to react to. Then I curate hard: I keep what’s accurate and usable and cut the rest. Every AI output passes a human quality gate before it counts.

How I collaborate with technical teams

I design so engineers can build without guessing intent. That means clear flows, explicit states, named edge cases, and specifications written to be implemented. I ask about data shape and constraints early, anticipate friction, and treat the build team as a design audience of its own.

How I balance speed, quality & judgment

Prototype to show, not tell — then let evidence decide. I ship the shortest correct version, instrument it, and iterate on the weak points. Speed comes from sequencing and reuse; quality comes from refusing to fake precision or skip the accessible path.

How I think about accessibility & ethics

Accessibility is designed in from the first layer (WCAG, Section 508, UDL) — built into templates and defaults so it scales. Ethically, I favor honest states over false confidence, avoid dark patterns, and keep a clear line between what a system knows and what it’s inferring.