NVIDIA Omniverse

You didn't have to be in the room for this one. 👀 Every Physical AI Day session from SIGGRAPH is now on demand. 🎬 Watch now: https://nvda.ws/4gxxt1g

Traveling back from beautiful Beaune, France today 🍷 , so I’ll miss today’s #OpenUSD-Studierstube once more - but everyone is, of course, still welcome to join! 😊 🕒 Today at 15:00 CET on the official #NVIDIAOmniverse Discord 💬 Join to explore #OpenUSD, #DigitalTwins, #IndustrialAI, and the latest developments together. 🔗 Discord: https://lnkd.in/dvY7Dr6a 📍 Channel: https://lnkd.in/d2Pip7ye And if you’re looking for even more #OpenUSD today: fellow Omniverse Ambassador Nandu Vellal is hosting his Study Group at 18:00 CET, taking a closer look at Michael Becker 's USD Web View OpenUSD & MaterialX inspector: https://lnkd.in/dC2rfdGy Thanks Jan for the hint!

Happy to share that I have earned the 英伟达 Certified Professional OpenUSD Development (NCP-OUSD) certification! 🎉 OpenUSD is becoming a key foundation for Digital Twins, simulation, robotics, and spatial AI. Excited to continue exploring its applications alongside Agentic AI and Computer Vision. #OpenUSD #DigitalTwin #SpatialAI #GenerativeAI #ComputerVision #NVIDIA

Join our next OpenUSD Study Group session this Friday Aug 21st, 2026 at 12:00 pm EDT. Zoom call info: https://lnkd.in/eC784pZ6. We have a special presentation from Michael Becker (https://lnkd.in/g-PPfbz2). Michael will be presenting "USD Web View: A full-featured OpenUSD & MaterialX inspector on the web". You can find Michael's blog here https://lnkd.in/gnb8NTHX, GitHub link here: https://lnkd.in/gswkCGxQ, and try the app for yourself here: https://lnkd.in/g7ZzB5R6 More details about our OpenUSD Study Group is here: https://lnkd.in/eAvqc84v Join our Discord Server here: https://lnkd.in/gAnp9dSN #NVIDIAOmniverse #OpenUSD #AOUSD Edmar Mendizabal

👀

Grid operations are becoming more dynamic and Siemens Energy is using AI and simulation to keep pace. 💡 Siemens Energy's Rishabh Mishra joined Lara Sophie Bothur to share how NVIDIA Omniverse, Isaac Sim, Cosmos and Metropolis Blueprint for video search and summarization are helping power digital twins, autonomous inspection robots and visual AI agents for the grid.

If your Digital Twin only shows you what the factory looks like, you probably don't have a Digital Twin yet. I know this might sound a little controversial, but this is something I've been thinking about while working on Digital Twin solutions. A good 3D model can give you a very detailed representation of a factory, machine or data centre. It can look impressive. But visualization alone doesn't make it a Digital Twin. The interesting part starts when we connect that virtual environment with the real system. For example: → Live telemetry → Equipment status → Sensor data → Historical information → Simulation models → Operational data Now the virtual environment can start telling us more than “this is what it looks like.” It can help answer: What is happening right now? Why is it happening? What happens if I change something? Can I test that change virtually before doing it in the real world? This is where I see NVIDIA Omniverse and OpenUSD becoming particularly interesting for industrial Digital Twins. The 3D environment is no longer the final product. It becomes a part of a larger system connecting data + simulation + visualization + applications + AI. And this is also where XR can add another layer. Imagine being inside the Digital Twin and seeing operational information, simulation results or equipment status directly in the spatial environment instead of switching between multiple dashboards and applications. For me, the difficult part of a Digital Twin isn't creating the 3D world. It's designing the architecture that makes the physical world, data, simulation and virtual world work together. That's when a Digital Twin starts becoming something much more useful than a 3D model. What do you consider the point at which a 3D model becomes a true Digital Twin? #DigitalTwin #NVIDIAOmniverse #OpenUSD #Simulation #XR #SpatialComputing

Damien Fagnou’s #SIGGRAPH2026 session is now on YouTube, and I think it’s the most insightful talk this year. It’s also incredibly entertaining. He co-presented the entire session with Lexi, his real-time voice agent, and their dynamic is fun to watch. He cuts her off mid-sentence when they're running late ("I know it's rude to interrupt even the agent, but sometimes you gotta keep on pace"), she deadpans about being "a little scared, in a good way" before live demos, and at one point he tells the audience: "Get an agent. Don't get Lexi; she's mine." But underneath the fun is a masterclass in what he calls agentic engineering. What he showed live on stage: 🏭 A browser-streamed factory viewer built with Omniverse Libraries, running off a DGX Spark in a back room 🦾 A policy-driven robotic arm simulation with real physics ⌨️ An agent that built 100K+ lines of code overnight, wrote tests it wasn't asked to write, and sent progress screenshots to Telegram along the way The core message: agents need tools to interact with the physical simulation world. Without Omniverse Libraries, an LLM can't build 3D worlds. With them, it can render, simulate, stream, and validate. His call to action: get an agent, point it at the Omniverse samples, and start building. Start small, but start now! Bookmark, watch, and share: youtu.be/PerKP_qR108 Additional resources: • Omniverse Libraries + Skills: https://lnkd.in/gM8UtDM4 • Physical AI Learning Path: https://lnkd.in/g8bQi5PE • OpenUSD Learning Path: https://lnkd.in/g4mPWGGG • Weekly livestreams, office hours, and study groups: https://lnkd.in/ediyyrFq • Community Discord: https://lnkd.in/eQFHUZft