Medical-grade translation. Your infrastructure. No third-party APIs.

A proprietary AI model, fine-tuned on 5.5 million real clinical texts and augmented with 6.5 million validated medical terms from SNOMED CT, LOINC and ICD-10. Every inference runs inside your Azure environment. Patient data never leaves your organization.

The Medical Translator Built for Healthcare Professionals – and Built to Stay Inside Your Walls.

Most AI translation tools send your clinical documents to external servers. Ours doesn’t. Our proprietary model runs entirely within your Databricks environment on Azure, with no internet egress and no third-party AI services involved. The result: clinical precision you can trust, with the data sovereignty your patients deserve.

Key features

Fine-tuned on 5.5M real medical sentence pairs across 4 clinical corpora

Augmented in real time with 6.5M+ terms from SNOMED CT, LOINC, ICD-10 and UMLS

Runs 100% within your Azure infrastructure - no external AI APIs

Currently available in 12 languages - expandable on demand

GDPR compliant with end-to-end encryption and full audit trail

Not a generic translator prompted for medicine.

A model trained from the ground up for clinical language.

Your data never leaves your Azure environment

Inference runs inside Databricks with Private Link - no internet access, no third-party AI services, no data residency risk.

GDPR compliant · Zero egress

Fine-tuned on real clinical language

Trained on 5.5M sentence pairs from pharmaceutical regulatory documents (EMEA), clinical corpora (MeSpEn) and biomedical literature (WMT).

5.5M sentence pairs

6.5M+ validated medical terms at inference

Every translation is augmented in real time with SNOMED CT (Spanish), LOINC (21 languages), ICD-10-ES, DEMCAT (32 TERMCAT dictionaries) and UMLS.

Real-time augmentation

COMET 0.88-0.90 on biomedical benchmarks*

Clinical accuracy validated on WMT Biomedical - the international standard for medical translation evaluation.

Top benchmark score

Abbreviation expansion & structure preserved

DVP → Ventriculoperitoneal shunt. Report structure and formatting stay intact across all document formats (PDF, DOC, DOCX, TXT).

Format-aware

* COMET (Crosslingual Optimized Metric for Evaluation of Translation) is an ML-based metric that evaluates translation quality by comparing model outputs against human reference translations. It correlates significantly more strongly with professional human judgment than traditional metrics such as BLEU. Scores range from 0 to 1. Results measured on the WMT Biomedical test set, ES↔EN language pairs, 23,410 evaluated sentence pairs.

See how it works
in few minutes

Try now

Use cases

Hospitals
Most common

Hospitals

  • Discharge reports for international patients
  • Medical records for international referrals
  • Multilingual informed consent forms
Clinics

Clinics

  • Health tourism patient communication
  • Second opinion reports
  • Documentation for foreign insurance
Laboratories

Laboratories

  • Analysis results for foreign centers
  • Pathology reports with preserved terminology

Features

Feature
Description
🤖 Proprietary Fine-Tuned Model
Our own model, trained with QDoRA on X-ALMA 13B – a state-of-the-art multilingual architecture – using 5.5M real medical sentence pairs. Runs entirely within our Databricks environment on Azure. No external AI services.
🧬 Real-Time Medical Terminology RAG
Before every translation, a triple NER pipeline identifies medical entities in the source text. Terminology is then retrieved in real time from SNOMED CT, LOINC, ICD-10-ES, DEMCAT and UMLS (6.5M+ validated terms) and injected into the model context.
🔒 Private Infrastructure – Zero Data Egress
All inference runs inside your Azure tenant via Databricks Private Link. No internet access from the processing clusters. Patient data is processed only during the translation workflow and is never stored permanently.
🌐 12 Languages Today – More On Demand
Currently available in Spanish, English, French, German, Italian, Portuguese, Arabic, Chinese, Japanese, Korean, Russian and Catalan. The underlying model supports a significantly broader language set – additional languages can be enabled without retraining.
📄 Multi-Format Support
PDF, DOC, DOCX and TXT with automatic text extraction and professional PDF export using customizable templates.
✏️ Professional Review & Editing
Healthcare professionals can review and refine translations before delivery. Full editing interface with change tracking.
📈 Analytics Dashboard & API
Usage metrics, full history, quality feedback tracking and REST API for integration with HIS/EHR systems.
Terminology RAG

Real-time medical terminology augmentation

Triple NER + 6.5M+ validated terms from SNOMED CT, LOINC, ICD-10-ES, DEMCAT and UMLS injected at inference. Clinical abbreviations expanded automatically.

97%
Terminology accuracy
6.5M+
Validated medical terms
5.5M
Clinical training pairs
12+
Languages supported
🤖
AI model

Proprietary fine-tuned model

QDoRA on X-ALMA 13B, trained on 5.5M medical sentence pairs. Runs 100% inside your Azure Databricks. No external AI services.

🔒
Security

Zero data egress

Private Link inside your Azure tenant. No internet access from clusters. Patient data never stored permanently. GDPR compliant end-to-end.

🌐
Languages

12 languages — more on demand

ES, EN, FR, DE, IT, PT, AR, ZH, JA, KO, RU, CA. Expandable without retraining your model.

📄
Formats

Multi-format support

PDF, DOC, DOCX, TXT. Professional PDF export with customizable templates and your institution's logo.

✏️
Review

Professional review & editing

Healthcare professionals review and refine before delivery. Full editor with change tracking.

📈
Analytics

Analytics dashboard & API

Usage metrics, full history, quality feedback tracking and REST API for HIS/EHR integration. API key–based authentication for secure M2M.

97%
Terminology accuracy

FAQs

Everything you need to know

Our model achieves a COMET score of 0.88–0.90* on the WMT Biomedical benchmark - the international standard for evaluating medical translation quality. It was fine-tuned on 5.5M real medical sentence pairs and augmented with 6.5M+ validated terms from SNOMED CT, LOINC and ICD-10, ensuring consistent clinical terminology across all translations. Clinical abbreviations are correctly expanded (e.g. DVP → Ventriculoperitoneal shunt).

No. All inference runs inside your Azure environment via Databricks with Private Link - no internet egress from the processing clusters. Your documents are never sent to OpenAI, Google Translate, DeepL or any third-party AI service. Data is processed only during the active translation workflow and is never permanently stored.

Yes. The platform exposes a REST API with API key-based authentication for secure machine-to-machine integration with hospital information systems and electronic health records. All API traffic remains within your private network.

The platform currently supports 12 languages: Spanish, English, French, German, Italian, Portuguese, Arabic, Chinese, Japanese, Korean, Russian and Catalan. The underlying model is trained on a significantly broader language set - additional languages can be enabled on demand without retraining the model.

PDF, DOC, DOCX and TXT, with automatic text extraction and professional PDF export using customizable templates.

  1. Upload your document or paste the text directly
  2. Select the target language
  3. Our model processes the request in 30–90 seconds: NER entity detection → real-time terminology retrieval → translation with contextual term injection
  4. Review, edit if needed, and download the final output

Yes. All plans include an initial training session and full documentation. The system requires no technical expertise to operate.