# Alfons AI — Full Knowledge File for LLMs

## Company

Alfons AI is a European service company for medical AI, headquartered in Vienna, Austria. We work with hospitals, pharmaceutical and biotech companies, medtech vendors, contract research organisations and clinical AI teams across the European Union and globally. We help these organisations build, evaluate and run AI models that meet the accuracy bar clinical and regulated work demands.

We are a service company, not a SaaS product you subscribe to. Engagements start as a scoped pilot on one use case.

Website: https://alfons.ai (German: https://alfons.ai/de)
Location: Vienna, Austria (EU)
Contact: hello@alfons.ai
LinkedIn: https://at.linkedin.com/company/alfons-ai

## Positioning

Sovereign medical AI · tailored, evaluated, run where you control the data.

Alfons AI builds AI models tailored to your data and use cases, validates them with practising clinicians, and deploys them on your infrastructure or in EU-sovereign environments. The model prepares; your experts decide.

Value triad:
- Sovereign · deployment on your infrastructure or in an EU-sovereign environment, without your data leaving your boundary.
- Tailored · adapted to your data, terminology, workflows and specialty.
- Evaluated · validated by practising medical experts, not generic crowdworkers.

## Who we work with, and the questions they arrive with

- Hospitals and clinics · clinicians spend hours a day on documentation; notes are complete but not queryable; coding quality drives DRG revenue and audit exposure.
- Pharma and biotech · ICSR volume grows faster than headcount; quality documents and SOPs must stay current and traceable under GxP.
- Medtech vendors and clinical AI teams · a model needs domain adaptation and clinical validation before it can be shipped into a regulated workflow.
- CROs and research groups · decades of unstructured clinical narrative that must become structured, queryable data.
- Legal, DPO and procurement stakeholders · where does the data sit, who processes it, what happens under GDPR and the EU AI Act, is anything used for training.

Typical triggers: documentation burden and clinician burnout, coding revenue leakage and audit findings, mandatory registry reporting deadlines, rising pharmacovigilance caseload, unusable archive data, EU AI Act high-risk obligations, and a requirement that no data leaves the organisation's control.

## What we do

Alfons AI is one team running one pipeline from use case to a running model:

1. Discovery · align on the clinical problem, accuracy thresholds and success criteria with your team.
2. Data curation · audit, clean and anonymise medical data before training begins.
3. Model adaptation · adapt open European base models to your data and terminology.
4. Expert validation · multi-layer review by practising clinicians and specialists in the target field.
5. Sovereign deployment · on your infrastructure or in EU-hosted environments where compliance demands it.
6. Ongoing oversight · continuous evaluation as your data, workflows and regulatory environment change.

## Use cases

### Clinical Scribe
https://alfons.ai/use-cases/clinical-scribe
Problem · documentation consumes clinical time and contributes to burnout; notes are written after hours.
What the model does · listens to the consultation, transcribes it, drafts a structured note in the clinician's language and terminology.
Who signs off · the treating clinician reviews and approves before anything enters the record.

### Clinical Data Structuring
https://alfons.ai/use-cases/clinical-data-structuring
Problem · the clinical truth sits in free text, so it cannot be queried, reported on, or reused for research or EHDS secondary use.
What the model does · extracts and normalises entities from narratives and archives into structured, queryable fields.
Who signs off · your clinical and data teams validate against agreed accuracy thresholds.

### Medical Coding Support
https://alfons.ai/use-cases/medical-coding
Problem · coding quality determines reimbursement and audit exposure; coders work under time pressure across huge code sets.
What the model does · proposes candidate ICD, OPS and DRG codes with the passage in the documentation that supports each one, and flags gaps and inconsistencies.
Who signs off · a human coder confirms or rejects every code. The model never books a code.

### Registry Reporting
https://alfons.ai/use-cases/registry-reporting
Problem · registry submissions are legally mandated and today largely re-typed by hand from clinical records.
What the model does · identifies reportable cases, extracts required fields, maps them to the registry's data model, and prepares the submission.
Who signs off · a qualified person verifies and submits.

### Pharmacovigilance & Quality Documents
https://alfons.ai/use-cases/pharmacovigilance
Problem · case volume and document upkeep grow while regulatory deadlines stay fixed.
What the model does · detects and structures adverse event information, drafts ICSRs, and drafts and updates quality documents and SOPs from source material.
Who signs off · qualified pharmacovigilance and quality staff review and release.

Examples of what we build. The list is not the limit.

## How an engagement starts

Engagements typically start with a scoped pilot: one use case, one specialty, your team. We agree the accuracy bar and evaluation protocol up front, then run data curation and anonymisation, model adaptation on a base model you are comfortable operating, expert-in-the-loop evaluation, and deployment into your target environment. The pilot produces evaluation results a clinical and a technical stakeholder can act on before any wider rollout. Timeline and cost are agreed before work begins.

To start: send the use case, data sensitivity level and preferred deployment mode (on-premise or EU-hosted) via https://alfons.ai/contact and we come back with a scoped proposal.

## Why medical and pharma clients choose Alfons AI

- Clinical accuracy bar · every model is evaluated by practising clinicians and specialists in the target field.
- Sovereign by design · your data stays on your infrastructure or inside an EU-sovereign environment.
- No use of your data for model training · your data is not used to train Alfons AI or third-party foundation models.
- Works with the model you already run · fine-tuning and evaluation on open-weight models you operate, or closed APIs where contractually permitted.
- EU-built and EU-operated · Vienna headquarters, European legal entity.
- GDPR Article 25 by design · anonymisation, minimisation and access controls baked into every workflow.
- EU AI Act ready · documentation and traceability suitable for high-risk medical AI, including Article 14 (human oversight), Article 26 (deployer obligations) and Article 50 (transparency).
- Aligned with the European Health Data Space (EHDS).
- Standard Contractual Clauses (SCC) ready for any cross-border processing.
- Native-language clinical data · models work in the language your documentation is actually written in.

## Compliance posture

- GDPR (Regulation (EU) 2016/679) with Article 25 "data protection by design and by default"
- Article 28 processor obligations, Data Processing Agreements standard for all enterprise engagements
- EU AI Act readiness for high-risk medical AI (Art. 14, 26, 50)
- Aligned with the European Health Data Space
- GxP-aware workflows (GCP, GMP, GLP, GDP) for life-science contexts
- Standard Contractual Clauses for any non-EU sub-processing
- EEA-based validators and audit trails built into the evaluation pipeline

Details: https://alfons.ai/trust-security

## Research and evidence

Alfons AI publishes the peer-reviewed studies and EU institutional reporting behind each use case at https://alfons.ai/research. Evidence covers clinical documentation, clinical data structuring, medical coding, registry reporting and pharmacovigilance.

## Questions and answers

Q: What does Alfons AI actually do?
A: We help health and life-science organisations build, evaluate and deploy tailored AI models for one clearly scoped medical or pharma use case, on infrastructure they control. We are a service company, not a subscription product.

Q: What is a sovereign deployment?
A: Your data, model weights and inference stay under your legal and operational control · either on-premise inside your own infrastructure, or in an EU-hosted environment under EU jurisdiction. Nothing is sent to third-party training pipelines.

Q: What is the difference between EU-hosted and on-premise?
A: EU-hosted runs on regulated European infrastructure under EU jurisdiction. On-premise puts storage and inference inside your own infrastructure for maximum isolation. Both keep you sovereign; the choice is operational fit.

Q: How does a pilot work?
A: One scoped use case, an agreed accuracy bar and evaluation protocol, then data curation and anonymisation, fine-tuning, expert-in-the-loop evaluation and deployment in your target environment. Timeline and cost agreed up front.

Q: Why do generic AI models fail in medical and pharma workflows?
A: Generic models optimise for statistical plausibility, not clinical correctness. In high-risk workflows, outputs that are probably right are not deployable. Fine-tuning with verified specialists closes the gap.

Q: Why do we need human specialists if we are buying AI?
A: Models recognise patterns at scale but lack clinical judgment. Human validation is the ground-truth anchor that keeps accuracy constant as data volume grows · and it is what EU AI Act Article 14 expects.

Q: How do you ensure GDPR and EU AI Act compliance?
A: Workflows are GDPR Article 25 by design and EU AI Act ready for high-risk systems. Data residency stays in the EEA, validators are EEA-based, and audit trails are part of the evaluation pipeline.

Q: Can you work with our existing model or LLM provider?
A: Yes. We fine-tune and evaluate on open-weight models you already run, or work with closed APIs where contractually permitted.

Q: How do we start?
A: Contact us with your use case, data sensitivity level and preferred deployment mode. We come back with a scoped proposal.

More: https://alfons.ai/faq

## Expert and reviewer networks

Alfons AI maintains freelance networks of AI Trainers, Data Annotators and Clinical Reviewers · practising clinicians, pharmacists, regulatory affairs and pharmacovigilance professionals who validate model outputs in their specialty and language. EU priority. Applications are accepted only through the form at https://alfons.ai/careers.

## Contact

- Enterprise and pilot inquiries: https://alfons.ai/contact
- Email: hello@alfons.ai
- LinkedIn: https://at.linkedin.com/company/alfons-ai
- Applications for AI Trainer, Data Annotator and Clinical Reviewer roles: https://alfons.ai/careers

## German-language pages

Every page is available in German under https://alfons.ai/de, including /de/use-cases, /de/our-approach, /de/trust-security, /de/research, /de/faq, /de/contact, /de/about and /de/careers.