Mohit Dutt

Problem

Single-AZ, hand-built environments drift from each other and have no recovery story. Development and production needed to share one codebase while differing in resilience and cost profile.

Approach

• Designed a highly available AWS environment with modular Terraform: multi-Availability Zone public and private subnets, an Application Load Balancer, ECS Fargate services, and a private RDS PostgreSQL instance.
• Hardened the platform with least-privilege security groups, deletion protection, and automated backup retention across separate development and Multi-AZ production configurations.
• Automated Terraform format, validate, and plan checks in GitHub Actions so drift and syntax errors surface on the pull request, not after apply.
• Authored SQL migrations, performance indexes, and PowerShell/Bash backup and recovery scripts to make the database tier operationally boring.

Outcome

One Terraform codebase produces both a lean dev stack and a Multi-AZ production stack. The database sits in private subnets with no public route, and every change is plan-reviewed in CI before it reaches AWS.

Problem

Manual deployments to App Service meant inconsistent environment config, secrets pasted into portals, and no gate between a bad commit and production.

Approach

• Built a full-stack MERN application with a multi-stage delivery pipeline, running unit tests, linting, and artifact builds in GitHub Actions before anything is packaged.
• Deployed the validated artifact to Azure App Service through Azure DevOps release stages, keeping build and deploy concerns separate.
• Automated resource provisioning and environment configuration with Azure CLI and PowerShell so environments are reproducible from scratch.
• Added secret masking in pipeline logs and pre-deployment hardening checks to keep credentials out of build output.

Outcome

Every commit is tested, linted, and built once, then promoted as an immutable artifact. Environment setup is scripted rather than clicked, and secrets never appear in logs.