AI intake website
Portfolio AI
An AI-native studio site that turns passive browsing into a live project briefing.
- ownership
- Design Engineer — designed & built solo
- timeframe
- 2 weeks
- stack
- Next.js, React, TypeScript, Gemini API
- proof
- public repo / live site / demo video
- outcome
- Live production site
decision
AI-native portfolio that answers questions, qualifies intent, and opens relevant proof.
build path
Designed in Figma → shipped in Next.js + React → exposed AI-readable docs and design-system routes.
scope
Build time
2 weeks
API integrations
Gemini API, Vercel server routes
System proof
Tokens, components, AI-readable docs
/design-system
Machine routes
portfolio.md, llms.txt, resume.json
outcome proof
Actual
Live production site
Source proof
Public Next.js repository
github.com/minwookshin/portfolio-ai
Recruiter proof
A live portfolio, source repo, and AI-readable system
Live
Production site
Public repo
Source-code proof
3
AI-readable routes
The problem
Static portfolios do not qualify intent
Most portfolio sites make visitors do the work: skim thumbnails, guess relevance, and hunt for proof. Portfolio AI behaves more like a studio strategist, answering questions, identifying what someone wants to build, and opening the most relevant project evidence.
Product proof
The site behaves like a small product system
Conversation qualifies the visitor's intent
Visitors can ask for an AI website, UX audit, product prototype, or case study. The response stays conversational, but it is backed by a clear intake path.
Project evidence opens deterministically
The chat can route people to the right case study without exposing implementation details. It feels like a dialogue, while navigation remains predictable.
What it demonstrates
Taste, system thinking, and applied AI in one surface
01
Live AI interface
Streaming responses, scoped system prompts, and guarded project routing are part of the actual portfolio experience.
02
Evidence-first navigation
The interface points recruiters and collaborators to concrete work instead of making the AI the whole product.
03
Machine-readable system
Portfolio content, resume data, LLM instructions, design tokens, and interaction rules are exposed as public routes.
Under the hood
System architecture
System proof
AI-readable interface primitives, not just a page
Roles instead of decoration
Color, type, spacing, radius, and motion roles are documented so new UI can be generated from the same quiet system.
Rules an LLM can follow
Component primitives, accessibility rules, reduced-motion behavior, and AI usage limits are exposed as public docs.
Turning a portfolio into an agency-style lead flow
Delivered a streaming AI interface that shows design taste, frontend craft, and applied AI in one live product.
Reframed the portfolio as a working studio demo: visitors can explore the work, ask about capabilities, and start a project brief from the same surface.