Migrated the entire platform for a US-based HealthTech startup from Heroku to Amazon EKS — eliminating scalability bottlenecks, reducing costs by 40–50%, and completing the production cutover in under 30 minutes with zero issues.
Context
The company, a dental aligner startup, was running their full platform on Heroku. Monthly spend was high, the platform faced frequent disruptions, and scalability had become a bottleneck. They needed a stable, cost-effective foundation built for future growth.
What I Did
Infrastructure Design
- Provisioned AWS Accounts with Organization Units — dedicated accounts for Development, Staging, and Production following security best practices
- Designed custom network architecture with centralized accessibility
Application Modernization
- Helped the team containerize all application services for Kubernetes deployment
Platform Build
- Deployed Amazon EKS as the compute platform
- Set up Bitbucket with self-hosted runners for CI/CD and ArgoCD for GitOps deployments
- Configured ALB with WAF for secure public API exposure
- Deployed CloudFront + S3 distributions for static site hosting
- Provisioned Amazon Aurora Serverless for the database layer
- Implemented observability stack with Prometheus, Loki, and Grafana
- Documented all infrastructure and standard operating procedures
Migration Execution
- Built a POC environment first and ran end-to-end validation tests
- Planned and scheduled migrations for each platform layer independently
- Set up database replication from Heroku to Aurora using chained replication
- In a scheduled maintenance window, cut over the entire platform in under 30 minutes — zero issues
Planning and execution by documenting every bit of the workflow is the key.
Results
| Metric | Before (Heroku) | After (EKS) |
|---|---|---|
| Cost reduction | — | 40–50% |
| Scalability | Frequent bottlenecks | Dynamic, auto-scaling |
| Cutover time | — | < 30 minutes |
| Downtime during migration | — | Zero issues |
Stack
Amazon EKS · Aurora Serverless · CloudFront · S3 · ALB · WAF · ArgoCD · Bitbucket CI · Docker · Prometheus · Loki · Grafana · AWS Organizations