╔══════════════════════════════════════════════════════════════════╗
║                                                                  ║
║   terraform apply && python janitor.py --dry-run                 ║
║   > Planning... 3 to add, 0 to change, 0 to destroy             ║
║   > Scan complete. 0 orphans. $0.00 waste detected.             ║
║                                                                  ║
╚══════════════════════════════════════════════════════════════════╝

I build cloud infrastructure and automation systems — the kind that prevents $1,700/month billing surprises, gives developers self-service access to AWS without touching the console, and makes every PR a cost hygiene checkpoint.

My work sits at the intersection of Infrastructure as Code, FinOps automation, and platform engineering. I write Terraform that enforces compliance by design, Python that detects cloud waste before it compounds, and CI/CD pipelines that make infrastructure changes as reviewable as application code.

Currently pursuing a DevOps internship while shipping production-quality cloud tooling.

AWS DevOps internship assignment — built to production standard

A complete multi-layer cost hygiene system for a fictional e-commerce client whose cloud bill grew from $400 → $2,100/month in one quarter.

What it does:

• Provisions NimbusKart's staging AWS infrastructure (VPC, EC2 × 2, S3, EBS) via Terraform modules targeting LocalStack — zero real cloud spend
• Python Cost Janitor scans for 4 orphan patterns: unattached EBS, stopped EC2 >14 days, idle Elastic IPs, and missing required tags
• GitHub Actions workflow runs on every PR — LocalStack as a service container, Terraform apply, 31 pytest/Moto unit tests, cost report uploaded as artifact, Markdown summary posted as PR comment
• Blocks PR merge if waste is found (exit code 1)

Key decisions documented:

• SSH rule uses dynamic block — completely absent from security group by default, not just restricted
• Delete mode re-fetches live tags right before deletion (race condition guard)
• Core layer imports nothing from providers — designed for GCP/Azure extension without rewriting

Stack: Terraform Python boto3 Moto LocalStack Docker GitHub Actions pytest

Self-service AWS infrastructure provisioning with real-time streaming logs

A production-grade platform that lets developers provision and tear down AWS resources (EC2, S3, Aurora, ElastiCache, Lambda) through a React dashboard — without touching the AWS console.

Architecture highlights:

• Zero-trust credential model: backend never stores secrets — Terraform registers credentials into AWS Secrets Manager, backend harvests them dynamically via the SDK
• Real-time orchestration: Socket.io WebSockets stream terraform init + terraform apply stdout line-by-line to a terminal UI — like Terraform Cloud but self-hosted
• Transactional rollback engine: provisioning failures trigger automatic state cleanup and DB reconciliation
• 5 Terraform blueprints: EC2 with 4096-bit RSA keypair + random password, Aurora PostgreSQL, ElastiCache Redis, Lambda + API Gateway, S3 with KMS-SSE
• CI/CD pipeline validates frontend + backend in parallel, then builds Docker images

Stack: Terraform Node.js Express React Socket.io Docker AWS Secrets Manager SQLite GitHub Actions

Event-driven microservices with streaming ML and graph analysis

A distributed fraud detection system designed for sub-50ms latency from transaction ingestion to risk decision.

Architecture:

• Edge Service (FastAPI + ONNX): XGBoost inference at ingestion, Redis velocity tracking, Kafka publish
• Graph Service (Python + Neo4j): GDS algorithms detect "mule activation" and "hub-and-spoke" fraud rings across account/device graphs
• Risk Engine: Aggregates ML edge score + graph signals + static rules → ALLOW / REVIEW / REJECT
• LLM Service: Generative AI explains why each transaction was flagged in plain English
• Real-time React dashboard via WebSocket streaming from final Kafka topic

Stack: Python FastAPI Apache Kafka Neo4j Redis ONNX Docker Compose React TypeScript

Written in Go

current_focus:
  - AWS DevOps internship at Code & Conscience (FinOps vertical)
  - Extending Cloud-Hygiene-Sentinel to scan GCP (provider abstraction layer ready)
  - Kubernetes fundamentals + EKS

interested_in:
  - Platform engineering roles
  - FinOps / cloud cost optimisation
  - DevSecOps pipelines
  - Multi-cloud IaC at scale

open_to:
  - DevOps internships and junior roles
  - Cloud infrastructure consulting
  - Open source collaboration on IaC tooling