Privacy-preserving infrastructure for clinical data
Active Crossing builds tools that let health organizations use large language models and share data for research without protected health information leaving their networks.
Our work
On-premises de-identification and re-identification for interoperability data
Most clinical data moves between systems in a few standard formats. Existing de-identification tools were built for free-text notes and do not handle these formats well. Our engine treats them as structured data.
- HL7 v2
- C-CDA
- FHIR R4
- Format-aware rules handle identifiers in structured fields and resources.
- A small, locally hosted language model finds identifiers in narrative text that rules miss.
- Everything runs on the customer's own infrastructure. The mapping between real patients and their surrogates never leaves the network, so results from a hosted model can be mapped back to the right patient locally.
- Utility is measured, not assumed. We test how well downstream tasks still perform on de-identified data.
This technology is currently in development.
Founder
Ryan Dick
Ryan is a Senior Integration Architect with more than 15 years in healthcare interoperability. He designs and runs the HL7 v2, C-CDA and FHIR integrations that move clinical data between health information exchanges, providers and health-IT systems.
- Statewide-scale health information exchange integration
- National exchange frameworks: TEFCA/QHIN, Carequality and eHealth Exchange
- Certified in Mirth Connect and InterSystems; integration engine upgrades and migrations since 2015
- Earlier career as lead developer on large-scale data systems in logistics and finance
- MBA and graduate certificate in health information technology