Large Language Models promise to revolutionize many aspects of life, but they also have many flaws, including the limitations imposed by the data they are trained on. These limitations are important in healthcare: people are understandably reluctant to allow their health data to be used to train models. Retrieval-augmented Generation (RAG) offers a possible solution, by providing the data with the question instead of during training.

To do this, health data needs to be in a form that can be consumed by a RAG system. FHIR is a great start, because it breaks the data into consumable chunks (resources), but those chunks are not easily consumable by RAG. This talk outlines how FHIR can be transformed to make it more accessible for RAG. The talk will show to how implement these techniques using John Snow Labs’ language models and libraries.

About the speaker

Sam Schifman

Principal Architect for Innovation at Availity

Sam Schifman has over 25 years of experience delivering software in finance, HR, education, no-code development platforms, and healthcare. For three years he was the Chief Architect at Diameter Health, until it joined Availity where he is now a Principal Architect for Innovation. In this role he focuses on the intersection of clinical and administrative data in healthcare. He has been at the forefront of delivering Upcycled data in FHIR, and other standards, to further interoperability. As part of that, he works with several HL7 Working Groups and Accelerators, including FAST and EHR/ Burden Reduction Working Group. He has experience with AI/NLP, having spoken at several NLP Summits and at HIMSS on the subject. Mr. Schifman continues to do research into how interoperability and AI can help reinvent healthcare and drive better outcomes for all.



Sessions: April 2nd – 3rd 2024
Trainings: April 15th – 19th 2024


Presented by