PulsHealth
PrivacyProtect Layer

Privacy-preserving middleware for health AI

Use it as part of the PulsHealth AI Health Agent or integrate it standalone into your own agent. On-device processing, minimum necessary data, and full user control.

How it works

A multi-step privacy pipeline that keeps users in control.

1

On-Device Analysis

An on-device LLM analyzes the query and identifies the minimum necessary clinical data to answer it.

2

Data Filtering

Only the relevant data is extracted and filtered. Full health records never leave the device.

3

Privacy Query Layer

When cloud LLMs are needed, data passes through aggregation, date fuzzing, and noise injection.

4

User Control

Users approve what gets sent, see an audit trail, and can opt for on-device only at any time.

Two modes of operation

Use PrivacyProtect as part of the PulsHealth ecosystem or integrate it into your own agent.

Built into the AI Health Agent
The PrivacyProtect Layer is deeply integrated into the PulsHealth AI Health Agent. Every query passes through the privacy pipeline automatically — no configuration needed. The on-device and cloud models have a conversation to solve your goals while protecting your health data.
Standalone middleware
Integrate the PrivacyProtect Layer as middleware in your own health AI pipeline. It sits between your agent and health data sources, handling on-device filtering, cloud query anonymization, and audit trails. Your agent gets the data it needs; users keep control.

Features

Every layer designed for privacy without sacrificing capability.

On-Device Processing
An on-device LLM analyzes queries and identifies the minimum necessary clinical data before anything leaves the device.
Minimum Necessary Data
Only the data required to answer a specific question is extracted. Full health records are never sent to external services.
Cloud Query Anonymization
When cloud LLMs are needed, data passes through aggregation, filtering, date fuzzing, and noise injection to help anonymize.
Audit Trail
Full awareness of what health data has been sent to which service, giving users and developers complete visibility.
User Control
Users stay in control at every step. They can use only the on-device model, and only when necessary will it suggest sending data off-device.
Middleware Architecture
Drop-in middleware for any health AI pipeline. Works as a standalone layer between your agent and health data sources.

Add privacy-preserving health queries to your agent

Join the waitlist for API and SDK access. Build health AI that users can trust with their most personal data.