OpenMed is a local-first healthcare AI toolkit for clinical natural language processing and privacy protection. It provides clinical named-entity recognition and HIPAA-oriented PII de-identification that run entirely on hardware you control, so patient data never has to leave your environment. With more than 5,700 stars on GitHub, it ships a library of over 2,200 medical models and runtimes for Python, Apple MLX, Android and the browser.
Healthcare teams increasingly need NLP, for extracting conditions, medications and procedures from notes, or scrubbing identifiers before data can be shared, but compliance rules make cloud APIs a hard sell. OpenMed packages fine-tuned medical models for exactly those jobs, with PII detection covering 35 model-backed languages and deployment options spanning servers, Apple-silicon machines, mobile apps and web browsers. It is designed for on-premise and sovereign-AI requirements from the start. The toolkit is maintained by a prolific open-source medical-AI developer whose models are widely downloaded, and the project emphasises practical deployment details: small specialised models instead of one giant model, predictable latency on local hardware, and SDKs that fit into existing hospital and health-app stacks. For organisations under GDPR, HIPAA or similar regimes, that combination of local execution and open licensing removes most of the procurement friction that usually blocks clinical AI adoption.
Free. The OpenMed SDK is open source under the Apache-2.0 license, including the model library, with no per-call fees.
Healthcare developers, hospitals and digital-health companies that need clinical NLP and de-identification while keeping protected health information inside their own infrastructure.