Privacy-Preserving Semantic Search in Genomic Variant Analysis: A Fully Air-Gapped Retrieval-Augmented Generation Architecture
Abstract The convergence of large language models (LLMs) and genomic medicine presents an unprecedented opportunity to accelerate variant interpretation, yet clinical deployment remains constrained by patient privacy mandates that prohibit transmission of genetic data to external services. We present a complete reference architecture for deploying Retrieval-Augmented Generation systems in air-gapped
