Portfolio · 2026

Lin Z.

Making complexity disappear.

Principal designer at the intersection of design systems and AI. I build the infrastructure that makes AI accurate: token architectures, semantic descriptions, and cross-platform mappings that let AI reason from your design rules instead of guessing at them. When the system is right, the whole organization ships faster.

Principal Product Designer · Kiva · Previously Upwork

Selected Work

Case studies

01
Color System
color/text/primary #223829
color/action/primary #276A43
Typography
Display
Headline 1
Headline 2
Body Text
Caption
Display44pxLight 300
Headline 126pxRegular
Headline 222pxRegular
Body16pxRegular
Caption14pxMono
font/display 44px / 300
Spacing
XS4px
S8px
M16px
L24px
XL32px
spacing/structure/M 16px
Architecture
Global Primitives
Alias Semantic
Mapped Components
3 layers AI-readable
90%+ AI Fidelity
KIVA DESIGN SYSTEM
Kiva · Principal Product Designer · 2025-2026

Building an AI-Ready Design System

AI tools hallucinate because design systems aren't built for machines. Rebuilt the architecture end-to-end with three token layers, semantic descriptions, and HTML and Tailwind mappings. Validated across three AI tools and scaled company-wide to 100+ teammates.

90%+ AI fidelity 3 tools validated 100+ teammates
View case study →
KvButton
×60Lend $25 Learn more Review
Repaid 98% Funded K
The Rule
figma said 3 button sizes code shipped 2 button sizes
code wins.
Impact
177 → 0
paint styles retired
ECOSYSTEM 2026
Kiva · Principal Product Designer · 2026

Code as the Source of Truth

The Figma library and the shipped code had drifted into telling different stories. Set one rule (when they conflict, code wins), retired 177 broken paint styles, and rebuilt ten core components token by token: a 60-variant button set, live-bound documentation, zero broken references.

177 styles retired 60-variant button set 0 broken references
View case study →
Skill Architecture
◆ SKILL.md
├─ reference/color.md
├─ reference/typography.md
├─ reference/spacing-layout.md
└─ reference/components.md
Zero Hardcoding
link text→text.action
btn primary→background.action
card title→title (Medium)
section gap→spacing.structure.XL
radius→base token
Impact
5 → 1
correction loops
KIVA AI SKILL
Kiva · Principal Product Designer · AI Pioneer · 2025-2026

Teaching AI to speak in tokens

5 correction loops per session, even with DTCG exports, Code Connect, and Figma Make. The format was wrong. Built a modular AI skill that reads once and never hallucinates: 91/100 score, 5→1-2 loops.

5→1-2 correction loops 91/100 skill score 100+ teammates
View case study →
04
Talent Scout desktop and mobile designs
73% conversion lift
40→16% no-show rate
3d→1d talent-match time
TALENT SCOUT
Upwork · Design Lead

Turning no-shows into converts

Flipped the flow order based on user mental models, lifting conversion 73%.

73% conversion lift
View case study →
05
Shoply V2 on iPhone
easy app good simple navigate clean modern follow feel nice easier change pretty clear colors bigger great flow
5-star App Store reviews
SHOPLY V2
Shoply · Lead UX/UI Designer

From frustration to retention

Owned V2 end to end: audit, redesign, testing, new design system.

31→82% retention
View case study →
06
Find Your Purpose design sprint prototype screens
UOPX · Design Lead

5 days to product clarity

Led a design sprint from vague brief to validated prototype.

5-day sprint
View case study →
07
The question
What did my money do?
The old answer
a ledger of zeros
142 moments, tagged
The new answer
a beginning, not a score
IMPACT PORTFOLIO
Kiva · Principal Product Designer · 2026

From 142 Moments to One Page

Every lender opens their account with one question: what did my money do? The old page answered with a ledger of numbers, and for new lenders, a ledger of zeros. Six usability sessions and 142 tagged moments turned it into a page that tells the story, for every lender, from the first loan on.

142 moments tagged 7 research-traced changes 0 zeros for new lenders
View case study →
About

Systems
thinker.

I'm a Principal Product Designer who builds design systems that work for AI as fluently as they work for humans. That means rethinking what "complete" looks like. Not just components and variables, but also semantic descriptions, cross-platform mappings, and documentation written for machines and humans alike. I've spent years closing the gap between design intent and what AI actually produces, then scaling that capability across entire organizations. My edge is finding the structural problem nobody has named yet, and turning the solution into everyone's new default.

Currently at Kiva · Previously Upwork · San Francisco Bay Area

LinkedIn → Email →
Contact

Let's
connect.

linzhao0324@gmail.com
LinkedIn