Skip to content

[ HELLO, WORLD ]

Based in Milan

Hi, I’m Mei.
I design AI-native
products that flow.

Product designer with 6+ years in UX and user research. I design B2B SaaS, AI agents and enterprise workflows: human-in-the-loop by design, built on design systems, and shipped in code.

Scroll to dive in

[ SELECTED WORK ]

Selected Work.

AI agents, B2B SaaS and enterprise systems, taken from research to shipped product, with the numbers they moved.

Experience

[ ABOUT ME ]

Curious by nature.
Designer by practice.

I design AI-native SaaS and enterprise products, with a background in interaction design at Sohu and user research at Smart Walnut. Most of my work is complex B2B workflows: agents that act for people, and the review, approval and control that keep those people in charge.

I work end to end: research and problem framing, interaction design, design systems in Figma and code, and front-end delivery with AI-assisted development.

Workflow observation and interviewsWhich problem is worth solving

Interviews, workflow observation, and product behavior show where people get stuck.

Printed interview scripts spread across a desk, with handwritten notes in blue ink.
Interview scripts with handwritten notes
Affinity board with five participant rows of notes clustered into themed columns.
Affinity map across five participant interviews
Journey and service mappingWhere the service breaks between people and systems

One picture of the whole chain: what a person does, what each system returns, and who picks up the next move.

User journey map with steps, actions, touchpoints, an emotion curve, and opportunities.
User journey with emotion and opportunities
Service map connecting a user, a tracking device, the app, a doctor, and the national health service.
Service map across device, app, doctor, and health service
Service blueprint connecting citizens, an AI chatbot, city projects, the municipality, and partners.
Service blueprint across citizens, city, and partners
Human and AI responsibility mappingWhat the agent does, where a person decides

Decision points, states, and exceptions, mapped with product and engineering.

Diagram: agent output, then a person reviews or dismisses, then product code saves the change.
Where the agent stops and a person confirms
Three-lane swimlane flow across end user, chat system, and support agent.
Cross-role swimlane: user, system, support agent
Design systemsWhat gets reused instead of rebuilt

Tokens, components, states, and usage rules, kept aligned between design and code.

Design system board with colours, typography, spacing tokens, buttons, icons, and cards.
Visual foundations and component vocabulary
Typography, semantic colours, spacing, and shape tokens taken from the ERP source.
Typography, colour, spacing, and shape tokens
Prototyping in Figma and codeWhether people understand it before it ships

The workflow made tangible early, to test how people read a recommendation, change it, and recover.

Wireframe board connecting landing, sign-up, matching, and interview screens with flow arrows.
Wireframe flow from sign-up to matching
A participant at a test table with the app on two phones, a feedback board, and a researcher taking notes.
Testing the app in person, with a board for how it felt
Behavioral evidenceWhat to change next

PostHog funnels, heatmaps, and session recordings show where people drop out.

Tools

Design & prototyping

AI & automation

Research & delivery

  • Figma
  • Storybook
  • Adobe CC
  • TouchDesigner
  • Claude Code
  • ChatGPT
  • n8n
  • Maze
  • Notion
  • Jira
  • PostHog
  • SQL

Contact

Get in touch

mei.chai@mail.polimi.it