Magnus EvensbergetTech lead & architect · Computas

This translation is AI-generated for now — a proper pass is coming.

Magnus Evensberget

Computas

MagnusEvensberget

Tech lead & architect

10+ years · full-stack · Oslo · NO / EN / 日本語

I'm a tech lead and developer, happiest taking an idea from a blank page to production-ready functionality — alongside a team that owns the result.

10+years as developer and architect
product teams taken from idea to production
  • Kotlin
  • Ktor
  • React
  • TypeScript

scroll · → for next

In brief

Ten years of cloud systems for organisations that can't afford to get them wrong.

For over a decade I've built systems for the Norwegian Police, NAV and Mattilsynet, and for private-sector clients like Wallenius Wilhelmsen and Pandion Energy — some in the public cloud, some on machines the client runs themselves. Today I lead the technology of a new product team at the Norwegian Police, building the next generation of interview tooling on locally hosted language models under strict security and privacy requirements.

I shape the architecture, choose the technology and write a lot of the code myself. I like being close to the detail, and I have opinions about design and user experience too.

Since 2022 I've carried people-management responsibility at Computas. That part of the job is mostly mentoring — following developers over time, helping them grow, and taking new ones in as the teams expand.

And I like the craft. Most weeks I'm reading up on or testing something new, and some of it usually finds its way into the day job afterwards.

Education

  • M.Sc. — Networks and distributed systems2011–2013University of Oslo · Department of Informatics
  • B.Sc. — Programming and networks2007–2011University of Oslo · Department of Informatics

Languages

  • 🇳🇴 NorskNative · C2
  • 🇬🇧 EnglishFluent · C1
  • 🇩🇪 DeutschBasic · A1–B1
  • 🇯🇵 日本語Learning · A1–B1

Capabilities

Architecture
  • DDD
  • Hexagonal
  • Event sourcing
Backend
  • Kotlin
  • Ktor
  • Spring Boot
  • Kafka
Frontend
  • React
  • TypeScript
  • Next.js
Platform
  • GCP
  • Kubernetes
  • Terraform
AI & LLMs
  • Local LLMs
  • Koog
  • AI-assisted development
Leadership & method
  • Team leadership
  • People management
  • Mentoring
  • Agile & TDD

Side projects

  • Language LearnerAn AI-powered app for learning Japanese — the Pimsleur method's strengths combined with real spaced repetition (FSRS). A single Compose Multiplatform client covers iOS, Android and desktop, with a React admin panel beside it. The backend is Kotlin and Ktor as a modular monolith, using Exposed and Flyway against Postgres. The model use is layered — frontier models where quality matters, local models via Ollama where they suffice, and models running straight on the device on iOS and Android.
  • poorboi.aiA Kotlin tool for spec-driven development that keeps the developer in control. Design is broken down into deliveries, tasks and subtasks, with the decisions and discoveries along the way written down beside them — so what survives is why something was done, not just what, and why the next thing should be done that way. It also gives agents like Claude Code and Codex the tools they lack on their own: running the tests, pulling structured logs, and verifying that build, test and lint are green before anything counts as done. It runs from a CLI with a React cockpit alongside, and the longer-term goal is for poorboi to drive the agent itself.
  • FinfinityMy own finance and algorithmic-trading platform. Kotlin and Ktor on the backend, built hexagonally around event sourcing, with React and TanStack on the front end. It collects market data, runs strategies against the history and optimises their parameters before anything is allowed to go live.
  • HomelabA GitOps-driven Kubernetes cluster on Talos Linux — the whole platform can be rebuilt from git.

Selected assignments

20262026–now

Politiet (PIT)

Team Avhørsflyt

Tech lead · Justice

Tech lead of a new product team building the Norwegian Police's next solution for planning, conducting and processing interviews.

  • From day one — the technical foundation for a brand-new team
  • Local LLM — transcription & report support
  • DDD + TDD — event-driven microservices
  • Kotlin
  • Ktor
  • React
  • TypeScript
  • Koog
  • Local LLMs
  • Domain-Driven Design
20242024–2025

Politiet (PIT)

Team Beslag

Full-stack · Justice

Full-stack delivery across web, mobile and backend in one of the Norwegian Police's first cross-functional product teams — modernising evidence-seizure handling.

  • 3 — surfaces — web, iOS and Android
  • Kafka — event-driven architecture
  • KMP — shared code across platforms
  • Kotlin
  • Spring Boot
  • React
  • Apache Kafka
  • Event Sourcing
  • Kotlin Multiplatform
20242024

Wallenius Wilhelmsen

Digital Supply Chain

Tech lead · Logistics

Helped stand up a brand-new logistics platform for global vehicle logistics — technology choices, data modelling and the developer platform on GCP. Partway through, the group took a strategic decision to swap to Azure and C#, and I saw the transition through. My last stretch there went on onboarding the new team and getting them up to speed, before I moved on to a new assignment.

  • Greenfield project
  • GCP — Cloud Run + Terraform
  • AI/ML — supply-chain optimisation
  • Kotlin
  • Next.js
  • Cloud Run
  • Terraform
  • Google Cloud
  • C#
  • Azure
20212021–2024

NAV

Employment schemes & providers

Full-stack · Welfare tech

New tooling for the ~1,000 providers who help people into or back to work — part of the effort to phase out NAV's ageing Arena system. The goal was to keep communication between the providers and NAV close, so that caseworkers and providers alike could follow each person up as well as possible. React micro-frontends inside NAV's own surfaces, and Kotlin services on GCP talking to the rest of NAV over REST, GraphQL and Kafka. Universal design was a requirement throughout.

  • ~1,000 — providers as users
  • 3 years — on the same product team
  • Arena — phasing out a legacy case system
  • WCAG — accessibility as a requirement
  • Kotlin
  • React
  • Apache Kafka
  • Google Cloud
20192019–2021

Pandion Energy

KAI Data Platform

Tech architect · Oil & gas

Pandion Energy wanted more value out of their geological and geophysical data, working with Computas and Google. Their owner Kerogen Capital had picked them as the first pilot for a global oil and gas data platform. The platform pulls data together from many sources, so the geologists can read across it and spot patterns they would otherwise miss. I was technical architect on a team spanning new graduates to experienced data scientists, and owned the portal they worked in day to day. During the pandemic the initiative had to scale down and was eventually put on ice, and I was the last developer left on the team.

  • Two full teams — grown from a small pre-project
  • GKE — hexagonal microservices
  • Terraform — a complete environment in minutes
  • Python
  • Java
  • TypeScript
  • Angular
  • Kubernetes
  • Apache Airflow
  • Terraform
  • Google Cloud
20172017–2019

Politiet (PIT)

The criminal-case chain

Developer · Justice

Three projects in the criminal-case chain: crime reports filed electronically into the police case system, case documents sent to the courts over Altinn, and maintenance of BL — the system 17,000 officers and prosecutors work in daily. I worked full-stack in Java, Spring Boot and Angular, and helped move the first services onto the police's new Kubernetes platform while they still had to run on the old one.

  • 17,000 — users on BL
  • Kubernetes — the first services on a new platform
  • Altinn — documents out to the courts
  • Java
  • Spring Boot
  • Angular
  • Kubernetes
  • Hexagonal Architecture
20132013–2017

Mattilsynet

MATS

Developer · Public sector

My first assignment out of university: developing and maintaining MATS, the Norwegian Food Safety Authority's inspection case system, which replaced around 30 older systems. Java and Oracle underneath, Angular on the newer surfaces — and test-driven development as a habit from there on. From 2015 I was team lead for one of the development teams.

  • ~30 — legacy systems replaced
  • Word → PDF — certificate generation rewritten
  • Lovdata — regulations pulled in automatically
  • Java
  • Oracle Database
  • Angular
  • Test-driven development