whoistylerdu.com

Personal portfolio website for Tyler Du — software engineer, curious mind.

Stack

Layer	Tech
Frontend	React 18 + Vite, React Router v6, react-markdown, CSS custom properties
Backend	Python 3.12, FastAPI, Mangum (Lambda adapter)
AI	Anthropic Claude Sonnet 4.6 — JD fit evaluation via /api/ask
Infrastructure	AWS Lambda, API Gateway, S3, CloudFront, SES, Route53, ACM
IaC	Terraform (state in S3, native S3 locking)

Local Development

Frontend

cd portfolio-website
cp .env.example .env.local   # set VITE_API_URL=http://localhost:8000
npm install
npm run dev                  # → localhost:5173

Backend

cd backend
pip install -r requirements.txt
uvicorn app.main:app --reload  # → localhost:8000

Deploy

Infrastructure (run once, or on infra changes)

cd infrastructure/terraform
terraform init
terraform apply               # requires terraform.tfvars with anthropic_api_key

Frontend

VITE_API_URL=https://api.whoistylerdu.com \
  bash infrastructure/scripts/deploy_frontend.sh

Or use the /deploy and /deploy-infra slash commands in Claude Code.

Note: Resume_Tyler_Du.pdf is not tracked in git. It lives at s3://whoistylerdu.com/Resume_Tyler_Du.pdf and is excluded from the frontend sync. To update it: aws s3 cp Resume_Tyler_Du.pdf s3://whoistylerdu.com/Resume_Tyler_Du.pdf --cache-control "max-age=86400" --region us-west-1

Architecture

whoistylerdu.com  →  CloudFront  →  S3 (React build, us-west-1)
api.whoistylerdu.com  →  API Gateway  →  Lambda (FastAPI)  →  SES / Anthropic API
                                                            →  S3 (resume PDF read)

Routes: /api/health · /api/contact · /api/ask

Ask Tyler (AI)

The /ask page lets recruiters paste a job description and get a Claude-powered evaluation of Tyler's fit. The agent is grounded in:

• Resume PDF fetched live from S3
• Project data from src/components/projects.js

Planned enhancements:

• MCP server integration with Google Drive (work samples)
• GitHub repository links surfaced as evidence
• Richer conversational context

Roadmap

• Phase 1 — UI overhaul (light + dreamy theme, Vite migration)
• Phase 2 — FastAPI backend + AWS infrastructure via Terraform
• Phase 3 — Agentic AI: Claude-powered recruiter assistant (/ask)
• Phase 3+ — MCP server (Google Drive, GitHub), richer agent context