I build AI systems for healthcare, language, and human wellbeing.
My work sits between research and production: evaluating medical AI, building open health infrastructure, developing language and speech technology, and turning clinical questions into usable software. I am a licensed clinical psychologist, PhD from KTH Royal Institute of Technology with a thesis on AI evaluation in medicine, Senior Research Scientist at AI Sweden, Senior Lecturer in Computational Linguistics at Uppsala University, and co-founder of Eir Space.
| Role | Focus |
|---|---|
| AI Sweden | Senior Research Scientist in the NLU team, working on language models and applied language technology for Swedish and European needs |
| Uppsala University | Senior Lecturer in Computational Linguistics, teaching and researching NLP, speech technology, and AI in healthcare |
| Eir Space | Co-founder, building privacy-first health tools that help people understand and use their own health data |
| Open Source | Medical AI benchmarks, health data tools, Swedish NLP resources, local-first apps, datasets, and models |
- Medical AI evaluation: benchmarks and methods for testing how language models reason in clinical contexts.
- Privacy-first health software: tools that keep sensitive health data local and useful.
- Speech technology for diagnostics: acoustic and language-based signals for clinical assessment.
- Swedish and Nordic language technology: models, datasets, and evaluation resources for underrepresented language contexts.
- Human-centered AI: clinical safety, harm reduction, explainability, and tools that fit real workflows.
Open-source infrastructure for privacy-first health AI. Eir Open brings together health data standards, medication lookup tools, local medical apps, clinical documentation experiments, and agent-ready modules for building patient-centered health systems.
Highlights:
- Local-first tools for viewing and working with Swedish medical records.
- Medication lookup resources for Swedish and US medications.
- Health.md, a lightweight standard for LLM-readable health records.
- Open medical scribe experiments that run locally.
- Swedish provider discovery and care navigation tooling.
A benchmark and evaluation framework for assessing large language models in the Swedish medical domain.
The work was published in Frontiers in Artificial Intelligence:
Moëll, B., Farestam, F., & Beskow, J. (2025). Swedish Medical LLM Benchmark: development and evaluation of a framework for assessing large language models in the Swedish medical domain. Frontiers in Artificial Intelligence, 8, 1557920.
An exploration of AI-assisted personal health management through journaling interfaces, where the interaction model matters as much as the model itself.
Related paper:
Moëll, B., & Aronsson, F. S. (2025). Journaling with large language models: a novel UX paradigm for AI-driven personal health management. Frontiers in Artificial Intelligence, 8, 1567580.
Audio processing and machine learning tools for automatic evaluation of the Pataka test, used in speech-language pathology and motor speech assessment.
My research focuses on the bridge between advanced AI methods and clinical utility.
Recent themes include:
- Evaluation of large language models in medical reasoning.
- Speech and language markers for neurological and psychological assessment.
- Harm reduction strategies for safe use of generative AI in healthcare.
- Synthetic clinical data and model evaluation.
- Human-AI interaction for personal health tools.
Selected publications:
-
High-accuracy prediction of mental health scores from English BERT embeddings trained on LLM-generated synthetic self-reports
Frontiers in Digital Health, 2026. -
Medical reasoning in LLMs: an in-depth analysis of DeepSeek R1
Frontiers in Artificial Intelligence, 2025. -
Swedish Medical LLM Benchmark (SMLB)
Frontiers in Artificial Intelligence, 2025. -
Harm reduction strategies for thoughtful use of large language models in the medical domain
Journal of Medical Internet Research, 2025. -
Automatic Evaluation of the Pataka Test Using Machine Learning and Audio Signal Processing
Acta Logopaedica, 2025. -
Multimodal capture of patient behaviour for improved detection of early dementia
Frontiers in Computer Science, 2021.
More: Google Scholar
I publish experimental models, datasets, and demos on Hugging Face, including:
- Clinical and Swedish-language LLM experiments.
- Swedish medical benchmark datasets.
- Speech and health-related datasets.
- Local model demos and comparison tools.
At Uppsala University, I teach and supervise in computational linguistics and language technology, including information retrieval and research-oriented language technology projects.
- Website: birgermoell.com
- GitHub: github.com/BirgerMoell
- Google Scholar: Birger Moëll
- Hugging Face: huggingface.co/birgermoell
- Uppsala University: staff profile
- AI Sweden: NLU team
- Eir Space: eir.space