Network Bio launched with $50 million in financing and announced a $30 million collaboration with Nvidia to build a cell-free RNA (cfRNA) foundation model for biomedical discovery. The startup said it will train models on large-scale tissue, blood, molecular, and longitudinal clinical outcome datasets sourced through partnerships with academic medical centers including Mass General Brigham, the University of Pennsylvania, and the University of Colorado Anschutz. Under the collaboration, Nvidia and Network Bio plan to develop a population-scale cfRNA model as a base for supervised models across oncology and non-oncology indications. The partners also described work to streamline the conversion of biobank access into model-ready datasets. Network Bio said it will use Nexus, a self-supervised transformer trained on cfRNA expression profiles, paired with Nvidia’s Parabricks nucleic acid sequence analysis software and modular tools from BioNeMo Recipes. The effort underscores how cfRNA is moving toward model-driven biomarker discovery and therapeutic development, with foundation models positioned as a way to generalize across disease contexts and improve signal detection from routine biological measurements.