Senior DevOps Engineer @ Ordaos Bio
Built Azure landing zones from existing subscriptions and an unreliable Pulumi prototype, then expanded the platform across GCP, AWS, and on-prem infrastructure for Kubernetes and GPU-backed compute.
- Expanded an Azure foundation into a multi-cloud landing zone, with Azure as the network hub for GCP, AWS, and on-prem connectivity
- Rewrote the inherited Pulumi code; delivered the first Azure development environment in 10 days, then built staging and production
- Provisioned and operated AKS, GKE, and EKS with Helm, Helmfile, and Kustomize; standardized application delivery through Argo CD
- Introduced Crossplane platform abstractions and Kubernetes admission webhooks to make provisioning and workload policy more consistent
- Built Kubernetes controllers and Go tooling for workload operations, including visibility into running images and their lifecycle
- Operated self-hosted HashiCorp Vault with Kubernetes integration, External Secrets, and sidecar injection for workload secrets
- Unified cluster telemetry with Prometheus, Loki, Tempo, Pyroscope, and Grafana for metrics, logs, traces, and profiles
- Created reusable Azure Pipelines and GitHub Actions workflows, including dedicated CPU and GPU runner capacity
- Reviewed cloud usage and identified savings opportunities through rightsizing, committed-use discounts, and reduced image egress
Key impact
- Delivered the first usable Azure development environment in 10 days, with staging and production following
- Connected cloud and on-prem environments through private hub-and-spoke networking, including site-to-site and point-to-site VPNs
- Identified infrastructure and image-distribution cost reductions; savings depended on workload and purchasing choices rather than a single fixed percentage