🧠 AI Digital Twin (Production-Ready)

This is a production-grade AI Digital Twin system built with AWS Bedrock + Serverless + Terraform + CI/CD.

Feel free to go to https://yujiangwu-digitaltwin.com/ to chat with me about my career and education.

---

🚀 Overview

This project implements a fully serverless AI application that:

• Serves a web-based AI assistant
• Maintains conversation memory
• Deploys infrastructure via Infrastructure-as-Code (Terraform)
• Supports multi-environment (dev / test / prod)
• Is CI/CD-ready with GitHub Actions

---

🏗️ Architecture

User (Browser)
↓
CloudFront (CDN)
↓
S3 (Frontend)
↓
API Gateway
↓
Lambda (FastAPI + Mangum)
↓
Bedrock (Nova Lite LLM)
↓
S3 (Conversation Memory)

---

⚙️ Tech Stack

Frontend: Next.js
Backend: FastAPI + Lambda + Mangum
AI: AWS Bedrock (Nova Lite)
Infra: Terraform + AWS (S3, API Gateway, Lambda, CloudFront, Route53, ACM)
DevOps: Git + GitHub Actions (next)

---

🔥 Key Features

• Serverless architecture
• Infrastructure as Code (Terraform)
• Multi-environment (dev/test/prod)
• HTTPS via ACM + CloudFront
• Persistent memory (S3)
• Global CDN delivery

---

🧠 What I Learned

• Terraform (providers, state, workspaces)
• IAM debugging (Route53, ACM)
• Serverless system design
• CloudFront + ACM constraints (us-east-1)
• Full-stack debugging (Frontend → Backend → AWS)

---

📦 Terraform Structure

terraform/
  main.tf
  variables.tf
  outputs.tf
  versions.tf
  terraform.tfvars
  prod.tfvars

---

🚀 Deployment

terraform init
terraform apply

Destroy: terraform destroy

---

⚠️ Notes

• ACM must be in us-east-1 for CloudFront
• Terraform is idempotent (no duplicate infra)
• Not transactional (partial infra possible)

---

📈 Future Work

• CI/CD with GitHub Actions
• RAG (vector DB)
• Tool-based agents
• Observability

---

🎯 One-line Summary

Production-grade serverless AI system using Terraform + AWS Bedrock.