Search Posts
Search blog posts by title, description, tags, or content.
Back to About
Infrastructure Modernization and Security Foundation
Day1 Company · Oct 2025 - Sep 2026
AWSGCPKubernetesArgoCDTerraformFirestoreValkeyOpenTelemetrywhisper.cppMetal
Tasks
- •Established Kubernetes operations through an ECS-to-EKS migration and built an ArgoCD-based GitOps deployment flow connecting builds, deployments, and completion notifications
- •Worked on AWS-to-GCP service migration, moving backend, web, and notification services to GKE and adapting deployment, cache connections, and authentication to the new environment
- •Migrated STT inference traffic from EKS GPUs to a shared Mac Studio to reduce GPU costs
- •Estimated baseline compute cost for two always-on GPU nodes at US$470–630/month using published minimum Seoul Spot rates and 730 hours per month
- •Built a whisper.cpp and Metal transcription server with a gateway compatible with the existing API, and switched production and non-production processing paths
- •Compared transcription quality using English and Japanese audio, and verified transcription, subtitle generation, and diagnostic reports for actual lessons
- •Used a WAF to block web attacks and managed security rules with Terraform, moving credentials into a secrets management service instead of storing them in code
- •Decoupled notification storage from DynamoDB through a storage interface and implemented a Firestore adapter and data migration tools
- •Designed for compatibility with existing queries, conditional updates, and scheduled-notification behavior, and wrote emulator-based regression tests
- •Implemented backfill, incremental synchronization, and pre-migration checks; disabled queue consumers and schedulers during migration to separate it from live notification delivery
- •Built an LGTM observability stack to reduce Datadog cost, connecting logs, traces, and metrics through MDC and trace IDs
- •Stabilized migrated services by adjusting JVM memory, resource settings, and log and metric collection paths for GKE
Related Posts
From ECS to EKS - Building an Operating Environment with kustomize and Gateway API
What I learned while migrating from AWS ECS to EKS and building an operating model with kustomize base/overlay, Kubernetes Gateway API, Karpenter node autoscaling, and blue/green deployment.
2026-03-12
Replacing Datadog with an LGTM Stack - OTel Sidecar with Grafana, Loki, Tempo, and Mimir
How I moved from Datadog to an open-source LGTM stack and built observability with an OTel Collector sidecar, noise filtering, MDC request tracing, and IRSA-backed S3 storage.
2026-03-17