TOOLS, LEARNING & SOFTWARE DEVELOPMENT
AI in practice.
From idea to use.
I build tools with AI, develop software around local models, and help people understand how to use them.
My projects connect technical experimentation with practical needs: working through a photo archive, completing tasks with local models, developing a virtual gallery and turning AI concepts into useful learning.
- Focus
- Tools, workflows & learning
- Work
- Personal projects & workplace learning
- Approach
- Build, test, review & improve
LOCAL AI WORKBENCH
Giving models
tools to work with.
A desktop application that connects local language and vision models with files, tools and tasks that take more than one step.
I work on the application around the model: context, model profiles, tool calls, saved progress and recovery when something fails. A central question is how the software can tell whether the work has actually been completed.
OPENCLAW · PERSONAL AI ASSISTANT
An assistant
within reach.
My OpenClaw installation runs on a Contabo VPS in Germany, giving me live access to an AI assistant from my mobile and other devices across networks.
It manages maintenance and updates across my servers and an extensive Obsidian archive containing around 20 years of medical records, clinical notes, medication history and blood-test results.
PHOTOSCANNER
Making a large archive
easier to describe.
PhotoScanner uses a local vision model to suggest titles, descriptions and keywords for my photo collection.
The collection contains around 80,000 photos. The tool saves progress, supports review and exports approved descriptions as XMP sidecar files for digiKam, keeping the original images unchanged.
AI AT WORK
Connecting understanding
with everyday work.
My work in training and development also shapes how I approach AI: people need a usable understanding of the technology and a clear reason to use it.
I have developed a self-paced course package for office and knowledge workers. It connects familiar chatbots to models, context, tools and agents, then helps learners define useful tasks and identify opportunities to improve a process.
AI-ASSISTED DEVELOPMENT
Working across
a complete project.
I also use AI coding tools to develop larger applications. Gallerum brings together a Unity gallery, shared visits, media archives and a browser-based collection workspace.
The work stretches from deciding how a feature should behave to implementing it, investigating failures, testing it and preparing releases. I guide the requirements and review the results through repeated use and testing. AI assistance is part of that development process; the project still needs decisions, integration and follow-through.
ONGOING EXPLORATION
Testing where
the ideas hold up.
Alongside these projects, I experiment with local model configurations, image and video generation, and turning images into 3D assets.
I have also developed the concept and an early code foundation for an AI-supported Danish news platform. The design separates source collection, evidence, drafts and editorial approval. It remains an early-stage project, rather than an active publication.
Across the work, I look at the complete task: output quality, time taken, errors, recovery and what a person needs to check. Model speed alone does not tell me whether a workflow is useful.
CONTINUE EXPLORING
Connected work.
Different applications.
Local AI Workbench
Models, tools and the application that connects them.
PhotoScanner
Local vision models applied to a real photo archive.
AI at Work
From understanding the chatbot to improving a workflow.
OpenClaw
A VPS-hosted assistant for server operations and personal knowledge.
IDEAS & PERSPECTIVES
Good ideas grow
through conversation.
I enjoy exchanging perspectives on people, technology and the things we learn by putting ideas into practice. If something here sparked your curiosity, let’s connect.
