I design, build, and operate infrastructure for stateful and distributed systems. My background includes data-centre, telecommunications, and blockchain environments. My work spans bare metal and Linux hosts, cloud-native platforms, networking, storage, observability, and release automation.
I take ownership of work from architecture through production. I turn operational needs into reproducible systems, define trust and failure boundaries, automate delivery, and make upgrades and recovery testable. I stay with incidents until the root cause is understood, including when the investigation crosses application, protocol, operating system, network, storage, or hardware layers.
I expect infrastructure changes to be reviewable and repeatable, with a tested recovery path. I use isolated tests and deployment gates for upgrades, migrations, and failure paths. When a production issue can be reproduced, I add a regression check.
I follow how cloud-native platforms and AI-assisted engineering are evolving. In AI systems, I focus on agent orchestration, controlled tool access, structured outputs, evaluation, and recovery. I treat model output as untrusted input and apply the same review, testing, and operational standards to AI components as to any other production dependency.
Most of my production work is private or employer-owned. The public work below provides concrete examples of how I design, build, and operate systems.
These repositories started from practical problems. Many are used directly as operational tools or as reusable foundations for production infrastructure kept in private repositories.
- Debian Kernel Current brings the newest authenticated Debian Sid kernel to Debian 13 without replacing its userspace. Its unattended release pipeline builds LLVM and ThinLTO packages, qualifies them in KVM, and publishes a signed APT repository.
- k8s-lab provides a self-hosted Cluster API platform on one Debian or Ubuntu host, using unprivileged LXC/LXD containers instead of virtual machines. It provisions declarative workload clusters and tests the complete deployment path.
- Remote SSH MCP lets coding agents work on remote Linux systems without a raw terminal. It exposes bounded operations over one fail-closed OpenSSH connection and moves large files outside model context with verified, resumable transfers.
- mini-pig is a reusable Ansible base for self-hosted bare-metal systems. It covers host setup, networking, VPN and proxy services, monitoring, and workloads, with a firewall policy that coexists with Kubernetes, Docker, libvirt, and Podman.
- Joplin tools connect a Joplin knowledge base to Markdown and coding agents. They provide safe two-way note sync, direct access through MCP and authenticated ChatGPT Actions, and deterministic migration from OneNote.
- FMI for Home Assistant brings Finnish weather forecasts, observations, lightning, and sea-level data into Home Assistant. This maintained fork adds tested migrations, failure isolation, compatibility checks, dependency review, and automated releases.
More personal projects and technical details.
My public upstream work includes production debugging, infrastructure automation, operational services, and SubQuery indexer fixes. Each entry is linked to the original issue or merged pull request.