Skip to content

Portfolio

Case Studies

What Interlusion has built itself, and the client projects from its founder's 25+ years in software engineering.

Interlusion

Own product, own platform

Bettles is Interlusion's own product, built AI-native. The platform it runs on is operated by one person, with AI assistance.

Bettles · Sports · Prototypes since 2018 · rewrite from December 2025 · live since June 2026

Built solo, AI-native

Role: Sole architect and developer

Result

UI, backend and infrastructure built by one person in about six months, live since June 2026.

Situation
Interlusion's own product needed a production-grade platform without a team.
Solution
  • Approach: Rewritten from scratch, steered through specification-driven development. The prototypes had about 10% of today's features; none of their code was kept.
  • Agents: Orchestrated coding agents across the whole lifecycle: architecture, implementation, tests, reviews, documentation.
  • AI in the product: An LLM commentator persona writes match previews and reviews; retrieval (RAG) and MCP keep it consistent across hundreds of generations.
Tech
  • Angular
  • Go
  • gRPC
  • NestJS
  • Redis
  • BullMQ
  • PostgreSQL
  • Keycloak
  • RAG
  • MCP
More about Bettles

Interlusion Platform · Own operations

Highly available Kubernetes, run by one person

Role: Sole architect and operator

Result

Runs Bettles in production and staging, and Interlusion's ERPNext. The ERPNext backups are restored automatically every week as a test.

Situation
Bettles and Interlusion's own business systems need production-grade hosting without an operations team.
Solution
  • Provisioning: A k3s cluster on Hetzner Cloud with a highly available control plane across three locations and autoscaling workers, provisioned with Terraform and Packer.
  • GitOps: Everything after provisioning runs through Flux CD, under a single-owner rule: each component is managed by exactly one tool.
  • Database: PostgreSQL on CloudNativePG with continuous WAL archiving to object storage.
  • Observability: Metrics, logs and traces flow into Prometheus, Grafana, Loki and Tempo.
  • Documentation: Architecture is documented as arc42, with ADRs and runbooks.
  • Operations: AI-assisted: Claude Code works through MCP servers for Kubernetes (read-only), Flux, GitHub, Prometheus and Alertmanager.
Tech
  • k3s
  • Hetzner Cloud
  • Terraform
  • Packer
  • Flux CD
  • Cilium
  • Traefik
  • cert-manager
  • Longhorn
  • CloudNativePG
  • Prometheus
  • Grafana
  • Loki
  • Tempo
  • GitHub Actions
  • SOPS
  • Claude Code
  • MCP
More about ERPNext

Before the agents

Client work

Projects Kersten Lorenz delivered for clients before the agent workflow existed. They stand for the founder's experience, not for work by Interlusion's agents.

IONITY · E-mobility and charging infrastructure · 2021–2025

Managing Europe's high-power charging network

Role: Coding architect for the frontend in an international SAFe organisation, plus backend services in his team

Result

The purchased system was replaced completely — on time, in the agreed quality, within budget. The organisation grew from two to around fifteen teams during the engagement.

Situation
IONITY managed its high-power charging stations with a purchased third-party product and was no longer satisfied with it.
Solution
  • System: A greenfield system on Kubernetes and AWS.
  • Frontend: Shaped and structured by Kersten Lorenz: UI foundation, design tokens, maps and charging-curve dashboards, large data tables built end to end with the backend.
  • Backend: Work on services for charging sessions, fleets, products and calibration-law-compliant records.
Tech
  • Angular
  • Taiga UI
  • D3
  • Leaflet
  • WebGL
  • Kotlin
  • Spring Boot
  • PostgreSQL
  • RabbitMQ
  • Kubernetes
  • AWS

Signal Iduna · Insurance · 2016–2017

From mainframe to web for the entire field sales force

Role: Software architect in the architecture team

Situation
The field sales force worked with a mainframe-based user interface.
Solution
  • Application: A new single-page web application in Angular.
  • Component library: Other departments built on it.
  • Infrastructure: A build and DevOps setup, and middleware between mainframe and browser.
  • Teams: Concepts, training and code reviews for the internal teams.
Tech
  • Angular
  • TypeScript
  • Java EE
  • Spring
  • Node.js
  • Jenkins
  • Webpack

Further references

  • smartlab

    2020–2021

    Charging infrastructure management, rebuilt

    As software architect, redesigned the frontend for managing charging stations and network operators and moved the team from a monolith to Docker and CI/CD on Azure.

    Tech: Angular · Java EE (WildFly) · Docker · Azure · Keycloak

  • Pflegix

    2019–2020

    A care platform with a new backend and billing

    As software architect and coach, extended the platform that connects carers with people in need of care: new NestJS backend, Stripe-based billing, frontend architecture, team training.

    Tech: Angular · NestJS · MongoDB · Stripe

  • Studienkreis

    2018

    Customer and contract management for a tutoring provider

    As developer and consultant, built the internal single-page application for managing customers and contracts.

    Tech: Angular · PHP · MS SQL

  • Huf

    2015

    Carsharing through the smartphone

    As developer, built a proof of concept for a fleet-management and carsharing platform in which the car talks to the smartphone over Bluetooth, mostly backend.

    Tech: Node.js · AngularJS · MongoDB

  • AGT International

    2015

    Operator interfaces for smart city and assisted living

    As developer and interface designer, linked ML-annotated traffic-camera timelines to the video and visualised fused sensor data for ambient assisted living.

    Tech: AngularJS · D3 · Node.js · MongoDB

  • VCE

    2013–2014

    A mobile app for administrative processes in healthcare

    As developer, built the frontend of a hybrid mobile app.

    Tech: AngularJS · Apache Cordova · Android

  • Fraunhofer IML

    2006–2013

    Matching logistics requirements to software

    As developer, part-time alongside his studies (about 20 hours a week), built a system that matches companies' requirements to warehouse, ERP and transport-management software — twice, in two departments on two technology stacks.

    Tech: Java EE · Apache Flex · .NET · MS SQL Server · XSLT · XSL-FO · XML Schema · D3

Let's talk

Kersten Lorenz will respond within 24 hours.

Get in Touch