Health Data and LLMs: Building a Secure LLM Playground for Healthcare on Google Cloud

Luis Alcántara and Marcos Dedeu present “Health Data and LLMs: An LLM Playground for Healthcare Customers,” focused on helping healthcare organizations use LLMs with sensitive data to optimize time-consuming workflows. They describe key healthcare data types (EMRs, clinical notes and research papers, genomic/imaging) and challenges: querying large corpora with RAG, reducing hallucinations via grounding and citations, managing prompt experimentation with a prompt library, and preventing proprietary data from leaving control through secure-by-design architectures. They explain the role of the FHIR standard for structured, machine-readable data and bridging FHIR with GenAI. Marcos demos a configurable, user-friendly playground that supports multiple models, RAG with citations, live FHIR retrieval and summarization, and prompt saving. They whiteboard and review a Google Cloud reference architecture using Cloud Run, load balancing, IAM, IAP, Cloud Armor, Model Armor, Sensitive Data Protection, AI datastores, FHIR store/Healthcare API, encryption/KMS, logging/monitoring, and VPC Service Controls. Ready to disrupt your industry? Google Cloud Consulting can show you how… https://goo.gle/4xQAEs9