Open Compute

Open Compute grew out of a frustration that anyone who has worked in health-tech knows intimately: clinical data is a mess. Patient journeys live trapped in unstructured text across fragmented systems. Meanwhile, the healthcare industry has spent decades building standards like FHIR to structure this data, but the gap between raw clinical narratives and clean, standards-compliant FHIR resources remains enormous.

The idea was deceptively simple: what if AI agents could read patient journeys the way a clinician does, and automatically transform them into structured FHIR data? Not through brittle rule-based parsers, but through genuine language understanding that can interpret context, resolve ambiguity, and map clinical concepts to the correct FHIR resources with the precision that healthcare demands.

Built as an open-source Python library at Jori Health, Open Compute uses a pipeline of specialized AI agents, each responsible for a different aspect of the transformation. One agent extracts clinical entities. Another maps them to standardized medical ontologies. A third constructs valid FHIR resources. The agents collaborate, cross-check each other, and produce output with full provenance tracking.

The decision to make Open Compute open-source was deliberate. Healthcare interoperability is too important to be locked behind proprietary walls. By releasing the library publicly, the goal was to give every health-tech developer access to the same AI-powered data transformation capabilities.

Open Compute represents a conviction that the hardest problems in healthcare are not medical — they are computational. The data exists. The standards exist. What was missing was the intelligence to connect them. That is what Open Compute provides: an AI-native layer that turns clinical narrative chaos into structured, interoperable data.