Platform
Clinical AI gateway
A gateway that sits between staff and AI models, masking patient information before it leaves your systems.
The Challenge
Clinical teams want to use AI for documentation and summaries, but every prompt is a chance to send protected health information to a service that shouldn't have it. Most practices either ban AI outright or don't know what's being shared.
How It Works
1
Requests go through the gateway.
Staff tools send AI requests to the gateway instead of directly to a model provider.
2
Patient identifiers are masked.
Names, dates, record numbers, and other HIPAA Safe Harbor identifiers are detected and masked before the request leaves.
3
Requests are routed by policy.
Each request goes to an approved model provider, with routing rules designed to keep sensitive work on providers covered by a business associate agreement.
4
Everything is logged.
Requests and masking decisions are recorded for review.
Built With
Kubernetes, Python, multiple AI model providers
What it's designed to do
- Let clinical staff use AI without sending patient identifiers to model providers
- Keep sensitive requests on providers covered by a business associate agreement
- Give the practice a record of what was sent where
Want something like this for your business?
Every build starts with an assessment, so we can adapt this pattern to how you actually work.