AI
Web
Google
HuggingFace
Local AI
Open Source
MedScribe
A privacy-focused clinical assistant that analyzes patient encounters locally using Google's medical AI models — no cloud APIs, no data leaving your device.
Project Highlights
1
Medical Speech Recognition
Convert clinical audio to transcripts using MedASR, optimized for medical terminology.
2
Multimodal Analysis
Analyze clinical notes and medical images (DICOM/X-ray) using MedGemma 1.5.
3
SOAP Note Generation
Automatically generate professional SOAP notes from patient encounters.
About the project
Inspired from latest MedGemma 1.5 release: https://research.google/blog/next-generation-medical-image-interpretation-with-medgemma-15-and-medical-speech-to-text-with-medasr/
MedScribe is a privacy-focused clinical assistant designed to analyze patient encounters locally without sending data to external cloud services. It combines Google's specialized medical AI models to process audio, text, and medical imaging — all on your device.
MedScribe enables healthcare providers to convert clinical audio recordings into transcripts using specialized medical speech recognition (MedASR), analyze clinical notes and medical images (DICOM/X-ray) using multimodal AI reasoning (MedGemma 1.5), and generate professional SOAP notes automatically.
All patient health information (PHI) remains on-device for HIPAA compliance. Only model weights are downloaded from HuggingFace — no patient data ever leaves your machine.
The application supports native M4 GPU acceleration on macOS with Metal and BFloat16 precision, and can also run fully containerized via Docker Compose for simplified deployment.
MedScribe is a privacy-focused clinical assistant designed to analyze patient encounters locally without sending data to external cloud services. It combines Google's specialized medical AI models to process audio, text, and medical imaging — all on your device.
MedScribe enables healthcare providers to convert clinical audio recordings into transcripts using specialized medical speech recognition (MedASR), analyze clinical notes and medical images (DICOM/X-ray) using multimodal AI reasoning (MedGemma 1.5), and generate professional SOAP notes automatically.
All patient health information (PHI) remains on-device for HIPAA compliance. Only model weights are downloaded from HuggingFace — no patient data ever leaves your machine.
The application supports native M4 GPU acceleration on macOS with Metal and BFloat16 precision, and can also run fully containerized via Docker Compose for simplified deployment.
Customize this for your use case? Please reach out to hello@appgambit.com
Key Features
- ✓Medical Speech Recognition: Convert clinical audio to transcripts using MedASR, optimized for medical terminology.
- ✓Multimodal Analysis: Analyze clinical notes and medical images (DICOM/X-ray) using MedGemma 1.5.
- ✓SOAP Note Generation: Automatically generate professional SOAP notes from patient encounters.
- ✓Local-First Privacy: All PHI remains on-device — no data sent to external cloud APIs.
- ✓Hardware Acceleration: Native M4 GPU support with Metal acceleration and BFloat16 precision on macOS.
- ✓Dockerized Deployment: Complete containerization via Docker Compose for simplified setup.
Tech Stack
Next.jsTypeScriptReactNodeJSFastAPIHuggingFaceGoogle MedGemmaGoogle MedASRDockerPostgreSQL
Keep Exploring
Want to keep exploring?
Here's another project you can jump into next.
Next project
Reccord
Reccord is a web-based, open-source screen recording application built with modern web technologies.
Web
Read next