AnyBio, a biosignal infrastructure platform for healthcare builders, and Medplum, an open-source FHIR-native developer platform, have announced a joint reference architecture for biosignal-driven care programmes.
The architecture is backed by a working integration over the FHIR Observation resource and a technical blog outlining how the two layers compose.
The combination gives healthcare builders an end-to-end deployable stack – from wearable device to clinician workspace – over a standards-based seam, instead of forcing them to retrofit a streaming pipeline into a clinical workflow product or vice versa.
The integration addresses a structural problem in healthcare engineering. Continuous biosignal data – the streaming output of wrist-worn and chest-worn wearables – and episodic clinical workflow – the FHIR-resource-shaped data model that electronic health records (EHRs), care plans and patient portals run on – require fundamentally different infrastructure underneath.
Streaming pipelines need different runtime, storage and consistency guarantees than transactional clinical workflow systems. Platforms that try to absorb both end up doing one well and the other badly, and builders pay for the weaker half.
AnyBio handles the continuous side, including Bluetooth Low Energy (BLE) ingestion from consumer and medical wearables, real-time signal processing and feature extraction, episode detection, governed AI agents and FHIR Observation emission with pre-mapped LOINC codes.
Medplum handles the episodic side through FHIR-native resource storage, CarePlan and Questionnaire scaffolding, patient portal and clinician workspace primitives, together with the consent, audit and authentication stack required for clinical deployment.
The boundary between the two platforms is the FHIR Observation resource, emitted natively by AnyBio and consumed natively by Medplum, with no proprietary translation layer between them.
The joint architecture is anchored to a detailed reference build, providing an integrated framework that helps developers bridge the gap between collecting a wearable’s raw BLE data and transforming those signals into clinical-grade, EHR-actionable records.
The reference models demonstrate how to build remote patient monitoring programmes from five patients to hundreds of patients, with the ability to enable patient-facing nudges two to three times a day through the Medplum web portal or the AnyBio iOS app.
The reference build also demonstrates how to establish real-time biosignal handling on the AnyBio side and clinical workflow presentation on the Medplum side without either platform stepping outside its specialist role. According to the companies, such a build would conventionally take 12 to 18 months and a full engineering team to develop from scratch.
Stephen Saine, Founder and CEO, AnyBio, said: “Continuous biosignals and episodic clinical workflow are two different engineering problems. Every platform that tries to absorb both ends up doing one well and the other poorly, and the builder pays for the worse half. We built AnyBio to be the biosignal plumbing and stay below the app layer. Medplum does the same thing for clinical workflow. Compose us over a FHIR boundary and a builder ships a real care program in weeks instead of spending 12 to 18 months building both halves themselves.”
Cody Ebberson, Co-founder and CTO, Medplum, said: “FHIR was designed for exactly this kind of composition. AnyBio emits FHIR Observation resources that can be consumed directly by Medplum, allowing shared customers to integrate biosignal data into clinical workflows without building custom translation layers. An Observation written by AnyBio lands in a Medplum FHIR server cleanly, and from there standard FHIR-based workflows can take over. Together, the two platforms demonstrate how specialized healthcare systems can compose through open standards.”
