I'm a medical doctor by training, based in Oslo, and I build software and automation projects that make everyday work a little easier.
My interest in software grew out of a practical problem. While working in healthcare, I saw how much time routine administrative tasks could take. I eventually built an automation system for a private outpatient clinic to help staff manage their daily workflow and reporting. Since then, I've been exploring other ways to use software to solve problems, from data pipelines and cloud integrations to AI-assisted tools.
I enjoy understanding how a process actually works, finding the parts that cause friction, and building something useful around them. My healthcare background has also made me attentive to accuracy, clear workflows and the people who will use a system.
-
Spreadsheet Workflow Automation — A real-world automation project for a private outpatient clinic. Built with Google Apps Script and Sheets to standardize entries, track workflow states, flag waiting cases and simplify daily reporting. The public version uses anonymized demos.
-
Podcast Automation — A Python-based pipeline for exercise science and nutrition podcasts. It combines transcript sources, checks disagreements, sends uncertain cases for human review and produces structured knowledge notes. The public demos use synthetic data.
-
House OS Automation — An ongoing project connecting everyday tasks, shopping and inventory through structured workflows. It includes Node.js integrations with Trello, Slack and Google Cloud, automated tests and a narrowly tested read-only assistant.
-
NutriValue — An early-stage project exploring how to connect grocery prices and nutrition information with Python, SQL and SQLite. The public portfolio includes a runnable synthetic demo, data modeling and validation checks—not live price recommendations.
-
AI-Assisted Development Workflow — An experimental workflow for building software with AI assistance while keeping project decisions, testing and review under human control. Still actively being developed.
- Automation and workflows: Google Apps Script, event-driven processes, reporting, APIs and integrations
- Programming and data: Python, JavaScript/Node.js, SQL/SQLite, data processing and validation
- Cloud and tools: Google Cloud, Git/GitHub, GitHub Actions, Slack and Trello integrations
- AI-assisted systems: transcription pipelines, model-assisted workflows, human review and testing
I use AI tools as part of my development process, but I don't treat generated output as proof that something works. I care about testing, tracing where data comes from, and being clear about what's operational, what has only been tested in a demo, and what is still an idea.
Most of these are independent, ongoing projects. Each repository has more detail on its current state, implementation and limitations.
P.S. Petricode is a nod to the Petri dish. :)