Open to product design, UX design, digital experience manager & web producer roles

Designer who thinks in systems and builds for the web.

I'm Dorothy Ly. I design digital products, websites and AI-enabled tools, and I understand the HTML, CSS and constraints that get them shipped.

Experience
5+ years in digital product & web design
Focus
Product design · design systems · AI workflows
Speaks
Figma · HTML/CSS · JS · CMS · SEO (basic)
Now
Prototyping an AI content assistant

Three projects, three different problems.

A shipped fintech rebuild, an AI tool in progress, and a product of my own. Hover, tap or tab through.

Some things I've designed.

Some things I've built.

Some problems I've tried to solve.

Where design meets AI.

I treat AI as a system with inputs, review and accountability, not a chat box bolted onto a product.

  1. 01 — AI workflows

    Take the repetitive work off people's plates.

    I map where a team loses hours, like research synthesis, content audits or first drafts, and design workflows where AI handles the volume and people handle the decisions.

    In practice: synthesising research transcripts into themes, and auditing content for freshness, readability and accessibility across hundreds of pages.

  2. 02 — AI UX

    Design for uncertainty, not just answers.

    Interfaces that show sources, signal confidence, make review fast and keep a human in control of anything that ships or gets sent.

    In practice: AI-suggested alt text, contrast fixes and ARIA notes that I review before components reach engineering.

  3. 03 — AI enablement

    Fit AI into how teams already work.

    Guidelines, prompt and retrieval patterns, and review steps that slot into an existing CMS or design process instead of living in a separate tool nobody opens.

    In practice: first-draft design system docs written alongside new components, and browser prototypes engineers can react to.

How I frame any AI feature AI step ■ · Human gate ■
  1. ProblemWhat's slow, costly or error-prone today?
  2. ContextGuidelines, data and examples the model must ground in.
  3. AIDrafts, classifies or summarises within set limits.
  4. The gateHuman reviewAccept, edit or reject, with sources in view.
  5. DecisionA person owns the call and the record of it.
  6. OutputShips into the tools the team already uses.
  7. MeasurementTime saved, edit rate, errors caught.

↺ Measurement feeds the next problem

I design with the build in mind.

I don't need to write every line of production code to design better products. Knowing how interfaces are built lets me design systems that are realistic, accessible, fast and easier to ship, and talk to engineers in their terms when it matters.

This site is the proof: hand-written HTML, CSS and vanilla JavaScript, built from the same tokens it documents.

  • Semantic HTML
  • CSS custom properties
  • Container queries
  • Canvas + WebGL
  • prefers-reduced-motion
  • Keyboard navigation
  • Progressive enhancement
this-page.jsonlive
frameworks
0
javascript shipped
—
dom nodes
—
dom ready
—
motion
—
effects
sleep when idle
Press I to see the type, spacing and grid specs.

AI Content Assistant

A small marketing team takes a campaign brief to an approved, SEO-checked, CMS-ready page, with a person deciding at every step.

Working prototype · not yet tested with users

  1. Strategy

    • AI proposes audience, intent, angle and structure
    • Sources listed
    • A person approves before any copy
  2. Write

    • AI drafts each section
    • Rewrite, shorten, expand
    • AI draft vs. human edit always visible
  3. SEO review

    • Checks computed from the text
    • Pass, warn or fail, with a fix
    • No overall score
  4. Approval

    • Every section approved
    • Final approval by a second person
    • Any edit withdraws approval
  5. Publish

    • JSON, HTML or Markdown
    • A person publishes
    • CMS adapters planned
  • Sources over guesses.Every strategy and section lists what it drew on. Quotes come only from approved proof, links only to pages that exist.
  • Rules, not a score.Named checks that pass, warn or fail tell a reviewer what to fix. A number only tells them how to feel.
  • Nothing ships unreviewed.The page can't be approved until every section is, and any later edit sends it back for approval.

A loop, not a line.

Each step ends with something concrete a team can react to.

  1. Discover

    Analytics, stakeholder interviews, the current build.

    → evidence, not assumptions
  2. Define

    Name the real problem and how we'll know it's fixed.

    → problem statement + metrics
  3. Explore

    Flows, IA and rough options, cheaply.

    → 2–3 directions
  4. Prototype

    In Figma, or in the browser when behaviour matters.

    → something to test
  5. Build

    Specs, tokens and pairing with developers.

    → shippable, accessible UI
  6. Measure

    Core Web Vitals, conversion, accessibility QA.

    → what actually changed
  7. Iterate

    Fix what launch revealed. Launch isn't the finish.

    → the next loop

Iterate feeds back into Discover

I like figuring out why something isn't working, then making it work better.

That's been true whether the thing is a checkout flow, a CMS editors are afraid to touch, a page that takes six seconds to load, or a team unsure where AI fits. I've spent 5+ years in digital product and web design, mostly where design, engineering, content and SEO overlap. That's where the interesting problems hide. My background spans product design, web platform work, CMS architecture and front-end development, so I can talk with engineering, marketing and leadership in their own terms.

I stay involved from research through to accessible, measurable, production-ready work, because that's where most designs either hold up or quietly fall apart. I'm curious by default. I read the source, ask the developer why, and prototype in the browser when Figma can't answer the question. Systems are what I enjoy most: design tokens, content models, workflows. Get those right and the screens get easier.

Lately that curiosity points at AI: not what it can generate, but how teams can trust what it generates.

Off-screen

  • HikingTrails over treadmills, every time.
  • PhotographyMostly light. Sometimes people.
  • TravelPlanning is half the fun. Hence Roamly.
  • PokémonType matchups were my first design system.
  • CoffeePour-over, slowly.
  • Running & lifting weightsSlow miles, heavy sets, good podcasts.

What I bring to a team.

Largest first. Each group has one line of proof, because a list alone proves nothing.

Product 01

  • UX strategy
  • Design systems
  • Information architecture
  • User research & usability testing
  • Prototyping

Made calling a specialist the first move on Kurv's paid-search page.

Web 02

  • Accessibility
  • HTML & CSS
  • Technical SEO
  • Performance
  • Responsive design
  • UX audits
  • JavaScript

Cut LCP on Kurv's blog by removing an image that wasn't earning its weight.

AI 03

  • AI UX
  • Human-in-the-loop
  • AI workflows
  • LLM interfaces
  • Prompt design
  • AI content systems

Designing confidence and review states for an AI content tool.

Tools 04

  • Figma
  • WordPress
  • Drupal & HubSpot
  • Git & GitHub (basic)
  • Adobe CC
  • GA4 & analytics

This site: Figma → hand-written code → GitHub.

Let's build something better.

Open to product design, UX design, digital experience manager and web producer roles. Email is the fastest way to reach me.