- Problem
- Requests were waiting seconds on the database, and the services around it were too coupled to scale independently.
- What I did
- Architected event-driven Java (Spring Boot) and Go microservices on RabbitMQ and Kubernetes (EKS), with AWS infrastructure in Terraform, GitHub Actions CI/CD, OAuth2/JWT for PCI-DSS, and Prometheus observability. Reworked PostgreSQL access with Redis in front of the hot reads.
- What happened
- 2M+ requests at 99.9% uptime. Queries down from 8.4 s to 420 ms, and incidents detected sooner.