Ambient AI
Clinical OS
A Gemini Live prototype exploring transcription, patient context, research grounding, and clinician review. The focus is on turning ambient information into something a clinician can inspect.
- Stage
- Exploratory prototype
- Tools and methods
- Gemini Live API, multimodal context, research grounding, safety checks, FHIR
View interface
The question
Can AI help carry the context of a visit without turning every observation into an unquestioned conclusion? This was an exploration of the experience around ambient clinical AI.
What I explored
I used Gemini Live to prototype transcription and multimodal context, alongside patient memory, research grounding, safety checks, and FHIR export. The workflow centers on making information available for clinician review.
The boundary
This is an exploratory product prototype, not a clinically validated system. It does not establish diagnostic accuracy, regulatory approval, or HIPAA compliance. The prototype should not be used for clinical decisions or with real patient data.
Workflow
- Capture context
- Structure evidence
- Review findings
- Prepare the record