Independent consulting

Kubernetes, infrastructure & AI engineering

I occasionally take on independent infrastructure engineering and architecture engagements.

I build Kubernetes-based AI infrastructure, from model deployment and inference routing to the platform services around them. My experience includes deploying models and implementing their routing layer, alongside extensive work on Kubernetes networking, controllers, and enterprise platforms.

I work across architecture, implementation, and production reliability, with the scope and ownership of a Staff or Principal engineer.

Tell me about your project

The work I take on

AI infrastructure & inference platforms: Design and build Kubernetes-based model serving with vLLM, KServe, and llm-d. GPU scheduling, inference-aware routing, autoscaling, and performance tuning, with attention to latency, throughput, and the cost of serving each workload.
Internal platforms & PaaS: Build self-service platforms on Kubernetes and OpenShift so teams can deploy and operate services with less manual work. Platform-as-a-service architecture, deployment workflows, tenant isolation, and the APIs and authentication that connect AI services to applications.
Kubernetes operators & control planes: Build operators, Kubernetes APIs, and Go controllers that automate infrastructure lifecycles. The difficult part is making reconciliation, upgrades, and recovery behave predictably when infrastructure fails or changes underneath them.
Networking across clusters: Design and implement how traffic enters and moves between Kubernetes environments. My focus is L4/L7 load balancing, Gateway API, Envoy, and the control planes that keep networking configuration consistent across clusters.
SRE, DevOps & production reliability: Establish SRE practices that fit your team: service objectives, useful observability, incident response, and recovery plans. Improve CI/CD, GitOps, and upgrade workflows across cloud, on-premises, and air-gapped environments.
Scaling, capacity & infrastructure costs: Find performance bottlenecks and unnecessary spend before adding more capacity. Review workload sizing, autoscaling, GPU utilization, and cloud costs; build capacity plans and infrastructure budgets around demand, reliability requirements, and measurable tradeoffs.

Experience behind the work

My background spans full-stack development, platform engineering, and production operations. That lets me follow a problem across application code, APIs, networking, and infrastructure, and work with the engineers responsible for each layer.

Over the past five years, I've built production infrastructure and helped design Kubernetes and OpenShift environments for enterprise customers. At Stakater, my work included platform engineering, operators, and deployment automation.

At Kubermatic, my focus has been multi-cluster networking, node lifecycle, and enterprise Kubernetes architecture. That includes leading KubeLB, maintaining Machine Controller / OSM, and contributing to Kubermatic Kubernetes Platform and KubeOne.

KubeLB · Project lead
Multi-cluster load balancing: Kubernetes APIs, centralized traffic management, and Gateway API integration.
Machine Controller / OSM · Lead maintainer
Worker-node provisioning, lifecycle management, and OS configuration across cloud providers.
Reloader & IngressMonitorController · Lead developer at Stakater
Controllers that automate workload rollouts when configuration changes and provision external uptime checks for ingresses.
Kubernetes LLM Inference Platform · Author
vLLM, KServe, and llm-d serving with GPU scheduling, inference-aware routing, tenant budgets, and performance observability.

How engagements work

You work directly with me, from the first technical discussion through delivery. We agree on the problem, the boundaries of the work, and what a useful result looks like before starting.

Review a design or an existing platform. Work through tradeoffs, failure modes, and operational risks, with written recommendations your team can act on.
Build a defined part of the system. Take an inference platform, controller, or networking capability from design through implementation and validation.
Work alongside your engineers. Help resolve architecture decisions and review the implementation as the team builds.

The handover matters: documented decisions, readable code, and enough context for your team to own what comes next.

Start with a little context.

A few sentences about your team and the problem are enough. I'll read your note and, if it looks like a good fit, follow up by email to arrange a conversation.