a cloud operator and software engineer based in Berlin. At SENEC I run large-scale Kubernetes platforms and build the infrastructure as code, GitOps delivery and automation that keep them secure, observable and up to date.
My background is in computational physics and data science: an MSc from the University of Oslo, peer-reviewed research in computational neuroscience, and production machine-learning work. In my own time I build open-source projects end to end, like Minnumst and recipify. I work in four languages: English, Icelandic, Norwegian and German.
Runs large-scale Kubernetes clusters and 50+ Ansible-managed servers, with GitOps delivery and automated, mTLS-secured maintenance.
Oversaw Azure Function App microservices integrating CRM and Business Central, built with Terraform, Terragrunt and modern CI/CD.
Built an LLM chatbot platform that simulates call-centre conversations on Django, Flask and Azure, and managed its frontend and UX teams.
Sales analytics of telecom data, and a natural language processing model that transcribes calls and captures the sales techniques that lead to sales.
A secure pipeline for automated maintenance updates across 50+ Ansible-managed, WireGuard-connected servers, built with strict access policies and secure networking. Hosts run a lightweight Go client for mTLS-secured maintenance checks, an API gateway schedules each host's maintenance slot, custom aptly-based repository mirrors serve the packages, and dashboards give an overview of host package lists and updates.
More about the role →A public archive, an editorial desk and an API that keeps every story revision, built end to end from the app to the hardened host it runs on. OpenTofu creates the Hetzner host, Ansible hardens it to the CIS benchmarks (scored with Lynis in CI), and Traefik with Authentik is the only way in. Releases are multi-arch images with an SBOM and build provenance.
Read the full story →
Turns about 1,900 mixed recipe files (Word, PDF, spreadsheets, scans) into 524 recipes, each checked word for word against its source and traceable back to it. 100% test coverage, with a live demo built by CI.
Read the full story →
Helped develop, package and formalize a tool that removes cardiac artefacts from spinal electrophysiology data, contributed to the open-source MNE-Python library.
Evaluated the kernel and hybrid approximation methods for computing local field potentials across a range of network conditions, simulating neuronal networks in Python with LFPy. Published in PLOS Computational Biology (2022).
corethe core stack, used at an advanced level
Hagen E, Magnusson SH, et al. "Brain signal predictions from multi-scale networks using a linearized framework."
Thesis: 'Evaluating Approximate Methods for Computing Extracellular Potentials from Neuronal Networks'. Grade: A.
Specialization in computational neuroscience, simulating neuronal networks in Python with LFPy.
Specialization in mathematics, statistics and computer science.
4D Space Numerical Workshop, Kobe University: simulated ionospheric effects on the NORSAT-1 satellite.
CaNoRock XVI, Andøya: simulated rocket trajectories to estimate drag coefficient and landing site.