I build AI assistants and automation for companies — and the business cases that decide whether they're worth building. Research Assistant at Aarhus University in Herning, Denmark, with a background in engineering and business development.
- Building: LUMA. A local-first personal AI assistant — and the small language model behind it, written from scratch.
- Researching: TraceableAI. An AI tool that helps researchers review literature — and shows its work, so nothing is made up. Headed for OSSYM 2026 in Berlin.
- Supervising: 3 MSc projects. Process diagrams from unstructured data, context engineering, and AI governance in a security firm.
- Learning: C# & .NET. Microsoft's main programming language and platform — already used to build AskEIVA and TraceableAI.
| Project | What it is | Built with | Code |
|---|---|---|---|
| Coursebot | Lets teachers build a chatbot for their course, no code needed | Laravel, Vue, PHP | private |
| AskEIVA | Support assistant that searches technical manuals and past tickets | C#, .NET 10, Clean Architecture | repo |
| TraceableAI | AI literature reviews that show where every result comes from | C#, .NET 10, Mistral Large | repo |
| LUMA | Personal desktop AI assistant with voice, news, camera and music | Python, PyQt6, Ollama | private |
| MIND @ LEGO | Search assistant over internal standards, used by 331 LEGO engineers | BM25, Similarity Search, Reranker | under NDA |
Mostly working with: Python · C# · .NET · RAG / GraphRAG · LLMs · Power Platform · SQL · PHP · JavaScript · Stakeholder management
Portfolio · LinkedIn · philip15@live.dk
The portfolio has a one-page quick view, and a chat assistant you can ask about my work.