NVIDIA

Partially latent flow-matching generative model for de novo atomistic protein binder design against protein and small-molecule targets, with experimental validation at million-design scale.

A training framework that grounds genomic foundation models with a joint-embedding predictive objective, learning to predict functional representations of masked DNA rather than reconstructing tokens.

Partially latent flow-matching model for joint generation of protein amino-acid sequence and full atomistic structure (backbone plus side chains) for proteins up to 800 residues.

A generative transformer foundation model for continuous glucose monitoring data, pretrained on 10M+ CGM measurements to forecast glycemia and stratify long-term health risk.

A family of codon-resolution language models trained on 130 million coding sequences from 20,000 species, revealing context-dependent codon grammar governing translation and mRNA stability.

A 3D vision-transformer foundation model for multimodal neuroimage segmentation, pretrained self-supervised on brain MRI from 41,400 participants.

NVIDIA/MONAI 3D medical image segmentation foundation model for CT and MRI, supporting automatic segmentation of 127 anatomical classes plus interactive point-prompt refinement.

Self-supervised vision transformer autoencoder pretrained on ~57,000 multi-contrast brain MRIs via masked image modeling for downstream brain tumor diagnosis.

A CLIP text-driven universal model for organ segmentation and tumor detection on abdominal CT, segmenting 25 organs and 6 tumor types with zero-shot extension to new categories.