Posts in category: Engineering

In a previous post, I described how we created a semantic layer over FHIR to support AI-driven capabilities such as semantic search and risk stratification. The response prompted a number of follow-up questions, many of them less about AI itself and more about the architectural decisions behind it: Why use FHIR narratives? Why are embeddings useful? What role does retrieval play? Why does chunking matter? And what exactly changes when agents are introduced into the architecture?
Healthcare interoperability often sounds straightforward—move data reliably between systems—but real‑world integration is far more complex, with differing models, semantics, and concepts. In a recent project, we faced this challenge in a sector with no existing interoperability standard, relying instead on a widely adopted operational dataset as our domain model. To bridge the gap with FHIR, we used StructureMaps and the $transform operation, enabling clean, scalable, bi‑directional mapping. This allowed us to deliver a robust transformation layer without building yet another bespoke engine.