CloudKeeper is building a Kubernetes optimization platform that keeps clusters running lean automatically right-sizing workloads and managing nodes without manual intervention. Two capabilities sit at its core:
- A recommender that analyses historical workload usage and applies corrected requests / limits automatically.
- Automated node management — provisioning, consolidation, and bin-packing for productive cluster utilisation.
This is a hands-on Go controller-author role: CRDs, controllers, admission / mutating webhooks, informer and reconcile loops, and a time-series layer for percentile-based sizing.
You'll work in a small engineering team building distributed systems that integrate deeply with the Kubernetes control plane. The work rewards someone who is genuinely curious about how Kubernetes behaves under the hood and who writes production-grade Go.
Responsibilities
Workload & Node Optimization
- Build the workload right-sizing recommender — pull historical usage from the metrics pipeline (metrics-server / Prometheus) and compute corrected requests / limits using percentile / histogram methods.
- Implement safe automatic application of those corrections to live workloads — staged rollout, guardrails, and rollback so a running workload is never destabilised.
- Build node-side optimization — provisioning, consolidation, and bin-packing — for efficient cluster utilisation.
Kubernetes Platform Engineering
- Design and build controllers, operators, and CRDs in Go using controller-runtime / client-go — informer caches, reconcile loops, the object lifecycle.
- Integrate deeply with the Kubernetes API to observe and act on cluster state at scale.
- Build the backend services and APIs (REST; gRPC where an agent / control-plane split calls for it) that the platform runs on.
Architecture & Scale
- Design for multi-cluster operation at scale.
- Hold platform services to a high bar for reliability,