I studied architecture before I wrote code. That training gave me one thing most engineers do not have: a framework for understanding how people navigate systems — not just how they click, but why they stop, turn back, or give up.
I apply that to e-commerce SaaS as a frontend engineer, and use AI to push that further — designing agents that intervene where filters fail, and systems that guide rather than overwhelm.
- Content distribution series — frontend lens on SEO/GEO → IA → analytics → recommendations
- AI engineering — Figma MCP, Spec-Driven development, production sub-agent workflows
| Project | What it is | Stack |
|---|---|---|
| yukiuix.com | Design engineer portfolio | Next.js, TypeScript, Tailwind |
| brooch-shop | AI shopping agent with tool calling | Claude API, TypeScript |
| post-agent-game | Human vs AI judgment comparison | React, Claude API |
| Yukiss (in progress) | AI-native stationery D2C | Next.js, Stripe, Vertex AI |
- Module Federation checkout rebuild → 30% of total orders via new flow
- Filter performance: CPU ~16s → ~8s via computation cache + render optimization
- i18n automation via Figma API, adopted cross-team
- Production support sub-agent: parallel Azure App Insight queries replacing manual triage
React · TypeScript · Next.js · GraphQL · Node.js · Vertex AI · Claude API · Mixpanel