NVDIA is significantly broadening its portfolio of open source AI models to enhance the fields of robotics, healthcare and autonomous systems. By moving beyond general language processing, these new model families such as Nemotron for digital agents and Cosmos for physical AI, are designed to help develops create systems capable of reasoning and interacting with the physical world.

A major portion of this deployment, is the focus on life sciences through the BioNemo platform. New tools like Proteina-Complexa allow researchers to design proteins for drug discovery, a process already undertaken by pharmaceutical companies to develop new therapies. Additionally, NVDIA has partnered with Google Deepmind, to expand the AlphaFold database, adding millions of new protein predictions to aid in drug discovery.

To further accelerate clinical research, NVDIA introduced nvQSP, a simulation engine that leverages GPU acceleration to model treatment scenarios. This engine offers an enormous boost over traditional CPU based methods, enabling scientists to simulate a vast arrray of dosages and patient variables much faster than previously possible, potentially streamlining the path towards clinical trials.

NVDIA open models, data and framework is available on Github and Huggingface and build.nvdia.com. Many of the models are also available on NVDIA Nim microservices, for secure and stable deployment on any NVDIA acceleration platform, from the edge, to the cloud.