Case studies

Building Azure Landing Zones, Then Expanding Across Clouds

Azure landing zones rebuilt from an unusable Pulumi prototype, with the first development environment delivered in 10 days before extending the foundation to staging, production, GCP, AWS, and on-prem.

Kubernetes & GitOps Delivery Platform

Delivery standardized across AKS, GKE, and EKS with Argo CD, Helmfile, and Kustomize. Progressive rollouts and Vault-backed secret sync gave teams a consistent release path without committing secrets to Git.

Unified Observability Platform

Metrics, logs, traces, and profiles unified across multi-cloud Kubernetes clusters in Grafana. Prometheus, Loki, Tempo, and Pyroscope support workload investigation, complemented by New Relic infrastructure and cluster instrumentation.

Internal Platform Tooling & Controllers

Developed custom Go CLIs, Kubernetes admission controllers, and image lifecycle controllers to automate operations and enforce policy as code.

Connected Mobility Infrastructure : Stellardian

A governed Azure delivery platform for Stellardian’s revamped connected-mobility and fleet workflows.

Connected Rental Platform : KIRRK

Terraform-managed Azure environments and governed AKS delivery for KIRRK’s connected rental and fleet platform.

Automated Tenant Provisioning : qomben

Terraform and Kubernetes delivery automation reduced tenant and environment setup from days to minutes.

Days to minutesTenant/environment provisioning

Building Azure landing zones after a failed Pulumi prototype

A failed Pulumi prototype rewritten into an Azure development environment delivered in 10 days, followed by staging and production.

10 daysto the first usable development environment

Rebuilding a brittle CI runner factory

How a stale image pipeline became a controlled Packer lifecycle for CPU and GPU Azure DevOps runner pools.

Moving a live Kubernetes platform to GitOps incrementally

How Kubernetes delivery moved from pipeline control to Argo CD through explicit, workload-by-workload handoffs.

Automating the lifecycle of cloud development machines

How identity-driven provisioning and scheduled lifecycle controls turned varied cloud workstations into a managed platform capability.

Reducing GPU scheduling waits from days to minutes

How generated ComputeClasses matched workload constraints to GPU and provisioning fallbacks instead of waiting on one capacity source.

Days to minutesObserved GPU scheduling waits

Giving pull requests preview environments on demand

How a pull-request label triggered an isolated Kubernetes application environment for focused review and testing.