Kubernetes Platforms
Multi-tenant Kubernetes platforms with self-service deployments for engineering teams.
Overview
A monolithic Java application serving a logistics platform had become impossible to scale — a single bug could bring down the entire system, and deployments required a 2am maintenance window every two weeks. This project decomposed the monolith into services running on a multi-tenant EKS platform.
The Challenge
The monolith had grown to over 1.2 million lines of code with no clear service boundaries. Any change required a full regression cycle (3 days), and a single memory leak in one module could exhaust resources for the entire application, taking down unrelated functionality.
Approach
- Mapped domain boundaries using Domain-Driven Design workshops with engineering leads — identified 14 candidate services
- Used the Strangler Fig pattern — new functionality built as microservices behind an API gateway, old code paths retired incrementally over 9 months
- Built a multi-tenant EKS platform with namespace isolation, network policies, and per-team resource quotas
- Implemented Horizontal Pod Autoscaler + Karpenter for workload-aware node scaling
- Service mesh (Istio) for mTLS, traffic shaping, and canary deployments between old and new services
Technology Stack
Results — Before vs. After
Deployment frequency
Mean time to recovery
Full regression cycle time
Infrastructure cost per service
Outcome
Deployments went from once every two weeks to 40+ per week across independent services. A failure in one service no longer takes down the platform. The team now ships features daily instead of in bi-weekly batches.
Facing a similar challenge in your organisation?
I work with engineering teams and businesses to design, build, and modernise cloud platforms — backed by a network of trusted specialists for execution at any scale.
Get In Touch