I'm a geek at heart. I started programming and hacking at 13, mostly with open-source software, taking things apart just to see how they work. Around the same time, I started caring about security and privacy.
That curiosity became my career. In telecommunications, I learned networking hands-on, designing networks and keeping them running myself, from fibre optics to the mobile core. In fintech, I learned how large companies approach security and process. Today I work as an SRE across data-centre, cloud-native, and blockchain environments. I design, build, and operate infrastructure for stateful and distributed systems, from bare metal to virtualization and Kubernetes clusters. My work also covers networking, storage, observability, and release automation.
I build tools that let coding agents work with remote systems and knowledge bases. I focus on controlled tool access, verifiable results, and recovery from failed operations, treating model output as untrusted input. Security is not only part of my job: I protect my own data the way I would protect a production system, with hardware keys, encryption, signing, and backups.
I take ownership from architecture through production. I turn operational needs into reproducible systems, automate delivery, and test upgrades and recovery paths. I stay with incidents until the root cause is understood, tracing failures across application, protocol, operating system, network, storage, and hardware layers. Reproducible failures become regression tests.
A few principles I rely on:
- The best way to solve a problem is to discuss it together.
- Check existing solutions first.
- The simpler the implementation, the better.
- A solution is only effective if it takes related tasks into account.
- If research isn't a challenge, it isn't research.
- Any task can be done in the best possible way, even if it wasn't planned like that from the start.
My day-to-day work spans many employer repositories. Some of them are public, but most of the code lives in private ones, so I don't list them here. Below are smaller projects I maintain myself. I sometimes use them as a foundation for solutions I build at work.
- Debian Kernel Current builds Debian Sid kernel packages for Debian 13 with Clang and ThinLTO, tests them in KVM, and publishes them through a signed APT repository.
- k8s-lab provides a self-hosted Cluster API platform with declarative workload clusters in unprivileged LXC/LXD containers on a Debian or Ubuntu host.
- Remote SSH MCP gives coding agents bounded tools for remote Linux administration and verified, resumable file transfers outside model context.
- SSH Wrapper is a typed Python library for sharing one OpenSSH connection across commands. Its heartbeat supervisor cleans up owned remote processes when their caller disappears.
- Elsewindow shows remote Linux application windows locally through Xpra over SSH. Rendering stays on the remote GPU; optional persistent sessions survive client disconnects.
- mini-pig provides Ansible roles for bare-metal hosts, networking, and services. Its firewall rules coexist with Kubernetes, Docker, libvirt, and Podman.
- Joplin tools provide two-way Markdown sync, MCP and ChatGPT Actions access to Joplin notes, 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 and release compatibility checks.
More personal projects and technical details.
I believe the open-source philosophy built the modern world. It's not perfect, but it really works, and I contribute back where I can.
My upstream work includes diagnosing a production HTTP/2 failure in HAProxy, fixing serialization and session handling in libnfs and fuse-nfs, and improving Wayland support, video decoding, and packaging in Xpra. Other contributions cover Polkadot deployment automation and SubQuery indexer fixes.