AI Lab
Built. Not just talked about.
Hands-on projects demonstrating applied AI security and agentic DevOps. Every project here is real, deployed, and publicly available on GitHub.
This website — Agentic DevOps with Claude Code
Designed and deployed entirely using Claude Code — Anthropic's agentic coding tool. Built with CLAUDE.md context engineering to teach Claude the project architecture, slash commands for reusable DevOps workflows, and a 3-layer safety model for infrastructure guardrails. Infrastructure provisioned with Terraform. Deployed automatically via GitHub Actions using OIDC authentication — no long-lived AWS credentials anywhere in the pipeline.
Deployment Architecture
GitHub repo → Actions (OIDC) → S3 → CloudFront CDN → amitwebsite.online
Terraform: S3 · CloudFront · ACM · Route 53 · OAC (Origin Access Control)
Security: HTTPS enforced · Security headers · No public S3 access · OIDC auth
Claude Code
Agentic DevOps
Terraform
S3 + CloudFront
GitHub Actions OIDC
ACM SSL
Route 53
LLM Threat Model Template
STRIDE-based threat model template for LLM deployments in regulated industries. Will cover prompt injection, data poisoning, model extraction, and agentic AI attack vectors aligned with MITRE ATLAS and OWASP LLM Top 10. Based on real threat modeling work at Lloyds Banking Group.
STRIDE
MITRE ATLAS
OWASP LLM Top 10
Regulated industries
Secure AWS Landing Zone — Terraform
Production-grade multi-account AWS Landing Zone with Zero Trust Kubernetes, IAM governance, SCPs, and security guardrails embedded in CI/CD pipelines. Based on real implementations at Standard Chartered Bank Singapore.
Terraform
EKS Zero Trust
AWS SCPs
Kyverno