Medalion provides an on‑premise AI platform for hospitals that lets medical staff build, train, and deploy machine‑learning models through a no‑code visual interface. The system processes clinical documents, lab results, and imaging reports locally, offering secure APIs for real‑time risk predictions, automated discharge summaries, and interpretable model explanations while ensuring patient data never leaves the facility.
Funding
Funding not disclosed
Founders
Product
Problem
Hospitals struggle to extract actionable insights from large volumes of clinical documents and structured data while maintaining strict data‑privacy regulations, leading to inefficient workflows and delayed decision‑making.
Solution
Medalion offers an on‑premise AI platform that enables medical staff to build, train, and deploy machine‑learning models without writing code. The system ingests hospital information systems, processes text, lab results, and imaging reports locally, and provides APIs for real‑time predictions and automated document summarization. Built‑in privacy features such as SSL/TLS encryption, on‑site processing, and RODO‑compliant anonymization ensure patient data never leaves the facility. Clinicians can generate risk scores, auto‑populate discharge summaries, and receive interpretable model explanations through a graphical interface, accelerating care pathways while preserving data security.
Target Audience
Primary customers are hospital administrators, clinical informatics teams, and specialty departments (e.g., cardiology, radiology, transplant units) seeking to embed AI into their care workflows.
Features
- No‑code visual interface for data import, model training, and deployment, allowing non‑technical medical personnel to create AI solutions
- Fully on‑premise execution; all data processing, model training, and inference remain within the hospital’s IT environment
- Pre‑trained domain models (e.g., MedalioNER for clinical entity extraction, AnonimiNER for GDPR‑compliant anonymization, LabExtract for OCR of lab reports) with >94% accuracy and SSL/TLS‑secured REST APIs
- Automated generation of discharge summaries and episode reports, reducing manual documentation time
- Real‑time risk prediction APIs (e.g., diabetes, cardiac event) integrated with existing Hospital Information Systems (HIS) via standard APIs
- Model interpretability tools (SHAP visualizations) to support clinical validation and trust
- Continuous monitoring and alerting dashboard for model performance and data quality