NVIDIA Healthcare

🚨 NVIDIA Medical Physics Simulation 🚨 A new open-source framework within NVIDIA Isaac for Healthcare helps developers model anatomy-device interactions, generate edge-case scenarios, and test in silico before hardware-heavy testing. Powered by NVIDIA CUDA, the framework unites classical physics (NVIDIA Warp and Newton) with generative AI dynamics (NVIDIA Cosmos). This powerful combination turns ad hoc simulations into scalable, GPU-accelerated infrastructure capable of running thousands of parallel environments to train the next generation of surgical robots. Developers have full access to the code, models, and weights to reproduce results, generate synthetic data, and build regulatory evidence on a GPU-native. 👉 Start building today nvda.ws/4bj19gE

The Open Secure AI Alliance is growing, with more organizations contributing expertise to help safeguard software and agents. nvda.ws/4pD8Fc5

Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community. During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion. That’s why we created the Open Secure AI Alliance.

NVIDIA's Open Models in 1 Diagram You probably know NVIDIA as the company that sells GPUs. But it’s also the world’s largest publisher of open AI models (@NVIDIAAI), with a lineup that goes far beyond chatbots. We wanted to understand how a company best known for hardware builds so many high quality models, and why it gives them away for free. So we spoke with @ctnzr, VP of Applied Deep Learning Research at NVIDIA. He walked us through the full lineup and the ideas behind it, summarized in the diagram below. 1. Reasoning: Nemotron is NVIDIA's family of LLMs that think before they answer. The line started in 2023 and the current generation, Nemotron 3, is a hybrid + MoE model that comes in three sizes: Nano for fast simple tasks, Super for harder planning, and Ultra for the heaviest reasoning. 2. World models: Cosmos predicts the next state of the physical world, not the next word. Cosmos 3 is one omni-model that can generate plausible video of a scene, reason about what is happening, and predict what comes next. 3. Robotics: Isaac GROOT is the first open humanoid foundation model. GROOT 1.7 takes camera frames and an instruction in, and outputs motor commands in real time. Its reasoning core is Cosmos Reason. 4. Production: safety, speech, and vision models. These are the building blocks that make AI usable in real applications. 5. Specialized domains: BioNeMo for biology and drug discovery, Alpamayo for self-driving cars that explain each decision, NVIDIA Ising for quantum computing, and Earth-2 for weather and climate modeling. Read the full breakdown here: blog.bytebytego.com/p/how-nvidia-b…

📢 𝗝𝗨𝗦𝗧 𝗜𝗡: $NVDA NVIDIA Launches Open-Source Medical Physics Simulation Framework for Healthcare AI 👉 𝗞𝗲𝘆 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀: ➤ 𝗡𝗩𝗜𝗗𝗜𝗔 launches 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗣𝗵𝘆𝘀𝗶𝗰𝘀 𝗦𝗶𝗺𝘂𝗹𝗮𝘁𝗶𝗼𝗻 framework for medical robotics. ➤ Open-source framework is part of 𝗜𝘀𝗮𝗮𝗰 𝗳𝗼𝗿 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲. ➤ Framework simulates anatomy-device interaction and synthetic training scenarios. ➤ Supports up to 𝟴,𝟭𝟵𝟮 parallel robot-training environments. ➤ Training time reduced from over 𝟱 𝗵𝗼𝘂𝗿𝘀 to under 𝟮 𝗺𝗶𝗻𝘂𝘁𝗲𝘀. ➤ Built on 𝗖𝗨𝗗𝗔, 𝗪𝗮𝗿𝗽, 𝗡𝗲𝘄𝘁𝗼𝗻, and 𝗖𝗼𝘀𝗺𝗼𝘀 technologies. ➤ Framework combines classical physics with 𝗖𝗼𝘀𝗺𝗼𝘀-𝗛 𝗗𝗿𝗲𝗮𝗺𝘀 generative AI. ➤ 𝗖𝗠𝗥 𝗦𝘂𝗿𝗴𝗶𝗰𝗮𝗹, 𝗝𝗼𝗵𝗻𝘀𝗼𝗻 & 𝗝𝗼𝗵𝗻𝘀𝗼𝗻 𝗠𝗲𝗱𝗧𝗲𝗰𝗵, and 𝗠𝗲𝗱𝘁𝗿𝗼𝗻𝗶𝗰 are adopting the framework. ➤ CMR contributed nearly 𝟱𝟬𝟬 𝗵𝗼𝘂𝗿𝘀 of anonymized clinical data. 👉 𝗘𝘅𝗽𝗲𝗿𝘁 𝗦𝘁𝗮𝘁𝗲𝗺𝗲𝗻𝘁: 𝗖𝗵𝗿𝗶𝘀 𝗙𝗿𝘆𝗲𝗿, Chief Technology Officer at CMR Surgical: "Open source models allow us to build on shared knowledge, accelerating responsible innovation and, ultimately, gives us the potential to deliver more consistent care and better outcomes for patients worldwide."

@JensenHuang @nvidia

🏥 MedTech Open Models Open models for physical AI and medical imaging - robot control, surgical simulation, segmentation, reconstruction, generation, and reasoning. huggingface.co/collections/nv… 🤖 Open-H Embodiment Open-H-Embodiment is a community‑driven dataset initiative building the open, shared foundation needed to train and evaluate AI autonomy models for surgical robotics and ultrasound. huggingface.co/datasets/nvidi… 🧬 BioNeMo Accelerated models for digital biology for design, understanding, and optimization huggingface.co/collections/nv… huggingface.co/collections/nv… huggingface.co/collections/nv… huggingface.co/collections/nv…

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-Weigh…

What does it actually take to turn an AI agent into a life sciences agent? NVIDIA BioNeMo Agent Toolkit combines open models, recipes, accelerated libraries, and NIM microservices so agents can run complex scientific workflows without managing the infrastructure themselves.

What if surgical robots could help to anticipate, plan and adapt in order to support #surgeons? CMR Surgical & @NVIDIAHealth are proud to showcase the next advance in #surgicalrobotics at @socroboticsurg SRS 2026 from today. #Versius #Healthcare #SurgicalRobotics #AI

250,000 downloads 🎉 Open-H-Embodiment just hit a big milestone. Developed with steering committee members from @JohnsHopkins and the @TU_Muenchen, and contributions from 50+ organizations worldwide, it is helping push the boundaries of embodied intelligence with real‑world robotic data. Explore it nvda.ws/4puobXR

📣 @bmsnews announced today it is deploying its second #NVIDIADGX SuperPOD, this one built with 8 DGX Vera Rubin NVL72 systems — making it the most powerful and energy-efficient AI factory in life sciences. ✅ Provide 10x performance per megawatt vs. previous infrastructure ✅ Give researchers access to a unified AI platform including NVIDIA BioNeMo Agent Toolkit for biological AI ✅ Managed by NVIDIA Mission Control software Learn how BMS is accelerating the discovery of next-generation medicines. ➡️ nvda.ws/4vyxsiN

Japan’s healthcare leaders are fusing cutting-edge AI with human care. 🤖 From autonomous surgical robots and AI-accelerated CT systems to agentic drug discovery platforms and virtual cell models, Japanese innovators are building a new healthcare stack on NVIDIA accelerated computing. This is Japan’s next generational leap in medicine. nvda.ws/4ymbi5Y

🌉 Building bridges between academia and industry for open biomedical data science Our AWS/NVIDIA hackathons trained 150+ researchers on real open datasets and produced 17 open‑source prototypes in knowledge graphs and federated genomics. If you're at #ISMB, come find @DCGenomics at poster B-046. Read preprint nvda.ws/4f4QyrV

As agentic loops in therapeutics become all the rage, it’ll be increasingly valuable to think about what score you’re maximizing. ipSAE is specifically good for prediction of positioning and alignment of residues at a protein-protein interface. But what if you care about alternatives like interface quality metrics (pDockQ, LIS), developability properties (hydrophobicity, viscosity, solubility), or immunogenicity/tox? We explored @NVIDIAHealth's Proteina-Complexa’s Composite Reward Model, which lets you customize the attributes and weights in a scoring function to guide agents toward the therapeutic goals that matter most to you. In this research article, we show how adjusting the weights changes the agent’s strategy to allow the model to adapt to different therapeutic design objectives. muni.bio/research/scori…

What’s the saying about the early bird and the worm? 🐦🪱 #NVIDIAGTC Berlin discounted pricing is available for a limited time only - join us for the latest in AI, accelerated computing, robotics, simulation, and more.

Check out resources👇 🔸Check out Fold-CP nvda.ws/4pexJWJ 🔸NVIDIA Transformer Engine (TE) Accelerates your Transformer Architectures nvda.ws/4h6ZSNk 🔸Learn how to integrate TE into your models from BioNeMo Recipes nvda.ws/3RykDHn 🔸cuEquivariance is open-source and available nvda.ws/44SVL02 (2/2)

Congrats to @biohub on ESMFold2 - a new protein structure model trained end-to-end on 256 NVIDIA H100 Tensor Core GPUs. The full NVIDIA CUDA-X stack made it possible: ⚡ cuEquivariance fused kernels for triangle multiplication - reduced compute overhead at the core of the attention mechanism ⚡ Fold-CP context parallelism - 4,000-residue proteins on 4× H100s; 6,500 residues on 16× H100s. Longer chains. Same hardware. ⚡ NVIDIA TransformerEngine FP8 low-precision training - faster iteration, lower memory footprint, no accuracy tradeoff Longer proteins. Faster folding. (1/2)

Learn more about how IPD built it on the NVIDIA stack: ▫️cuEquivariance → geometry-aware ops at the model's core nvda.ws/3QMxQw3 ▫️BioNeMo Agent Toolkit → IPD among the first testers nvda.ws/3TuzZx5 ▫️MMseqs2-GPU → GPU-accelerated sequence search nvda.ws/4hc5c1Q

⚡ The @UWproteindesign is accelerating the next generation of protein design. Using NVIDIA cuEquivariance and TensorRT, IPD sped up RF3 structure prediction from 5.8s → 2.5s on a 256-residue protein, a >2× speedup, right at the geometric core of their model. 🧬 See how nvda.ws/4pmFOc4 (1/2)