NVIDIA Physical AI Community

Welcome to the Nebius x NVIDIA Global AI Hackathon

Build a working AI system on open, independent infrastructure, and own the whole thing. The NVIDIA x Nebius Global AI Hackathon is for AI builders, ML engineers, founders, and anyone shipping practical, high-performance AI. You'll build with NVIDIA open source models served on Nebius through Nebius Token Factory, on high performance GPU infrastructure. Experiment with new ideas, push the boundaries of open source AI, and submit a project for a chance to win prizes and recognition.

Join fellow builders for technical onboarding, casual workshops, team formation, focused build time, and feedback as you participate in the Global AI Hackathon. Find your nearby city/event here!

The Challenge
All submissions must run on either Nebius Token Factory or Nebius AI Cloud and use at least one NVIDIA open source model. Everything else is up to you.

What to Build
Whether it is an app, agent, or something entirely new, your project should fit into one of the following tracks:

• Coding and Agentic Engineering Track
  • Build coding agents and developer tools: agents that write, run, and test code in Token Factory Sandboxes.
• Best Apps and Agents Track 
  • Build any app or agent someone would actually use, from a productivity tool or copilot to a workflow that runs itself. Power it with Nemotron models on Nebius through Token Factory. Reach for Nemotron 3 Ultra when you need serious reasoning, and let Nano or Super handle the fast, everyday calls, so your app stays responsive and your credits stretch further. We encourage you to deploy your application with Nebius Serverless Endpoints, or use Nebius Serverless Jobs for background processing and asynchronous workflows, though neither is required.
• Personal AI Track
  • Build an always-on, private assistant that works for you while keeping your data under your control. Give it persistent memory, reusable skills, access to the tools and information you choose, and the ability to carry out tasks across your daily workflows. Use at least one NVIDIA open source model, and use tools such as NVIDIA NemoClaw, OpenShell, Hermes Agent, and Nebius Serverless to bring personal AI to assemble, secure, and run your own personal AI system.
• Physical AI Track
  • Build embodied and edge agents that sense and act in the real world: robotics, IoT, and on-device intelligence, powered by Nemotron, GROOT, Cosmos and Sonic models and coordinated through an agent runtime. Use Nebius Serverless Jobs to run simulations, generate synthetic data, evaluate robot policies, or process sensor data at scale, and Nebius Serverless Endpoints to serve real-time inference when needed. Your demonstration video must include at least a 1-minute clip showing the physical hardware/robot actually operating, or — if your project has no physical hardware component — the key application modules in action.

What to Submit

• A working project. Build something with NVIDIA Nemotron or other NVIDIA open source models on either Nebius Token Factory or Nebius AI Cloud that meets the track requirements.
• A category. Pick the track that is the best fit for your project.
• A project description. Tell us what you created, why you created it, and how it works.
• A working demo. Provide a URL to a working demo, hosted application, or test build of your project.
  • Not required for Physical AI submissions.
• A demo video. Upload a 3-minute or shorter public YouTube video showing your project working, with audio covering how you used Nebius Token Factory and NVIDIA Nemotron or other NVIDIA open source models. 
  • Physical AI submissions should include at least one minute of footage showing the physical hardware/robot actually operating, or — if your project has no physical hardware component — the key application modules in action.
• A public code repository. Submit a publicly accessible code repository URL (GitHub, GitLab, or Bitbucket) for judging and testing. The repository must include an open source license (such as Apache 2.0, MIT, or MPL 2.0), visible at the top of the repository page.
  • Include a README with setup instructions and clear guidance for running your project.
  • Make sure to highlight how you've used NVIDIA Nemotron or other NVIDIA open source models, where Token Factory accelerated your workflow, and any other Nebius Tools or Services used.
• Feedback on Nebius Token Factory, AI Cloud, and any NVIDIA tools, models, or technologies used.
• If your project existed before the Submission Period, include a written explanation of what was significantly updated during the Submission Period.
• If you participated in one of the IRL Builders & Brews events, choose the city you attended for a chance to win the City Winner Award.
• Review the complete Official Rules. 

The NVIDIA GTC AI conference is coming to Berlin, bringing together developers, researchers, and industry leaders to go deep on AI infrastructure. Sessions will span the entire five-layer AI stack, from energy, chips, and infrastructure to open models and physical AI applications empowering every industry and organization to build AI on their own terms.

See CEO Jensen Huang's live keynote at the iconic Tempodrom, then join us at STATION-Berlin for dynamic sessions, hands-on labs, and exclusive networking.  

Join NVIDIA and Mode Maison for a casual coffee meetup during GTC Berlin! Connect with fellow developers, researchers, and NVIDIA experts to exchange ideas and explore what's happening across Physical AI, simulation, and beyond.

We'll be joined by Mode Maison, an independent research lab working at the intersection of physics-based simulation, AI, material science, design, and manufacturing, and the creators of Total Material Appearance Capture (TMAC).

Whether you're attending GTC Berlin or just in the area, everyone is welcome! Come grab a coffee, meet the community, and join the conversation.

Hear from NVIDIA CEO Jensen Huang live on stage at the iconic Tempodrom to kick off GTC Berlin. Arrive early to catch the GTC Live pregame show for an insightful discussion on the latest in AI infrastructure with industry leaders.  

GTC Berlin Keynote Information Here 

The keynote is over. Let’s talk about what it means for developers. Join NVIDIA experts and the community for an after-hours conversation from GTC Berlin, featuring first reactions, a look at the day’s announcements, and perspectives on open models and agent safety. Interact live with special guests from Hugging Face and Pollen Robotics.  Hear what’s generating excitement across the event, test your knowledge in an interactive quiz, and enjoy a few community moments along the way. Bring your questions and help shape the discussion about what comes next.

Welcome to the Isaac Sim & Isaac Lab Study Group, your virtual learning space for robotics simulation and development!

Led by NVIDIA Omniverse Ambassador Scott Dempsey, this interactive study group is a place to learn, build, share, and grow together with the robotics community.

Whether you’re just getting started or already building with NVIDIA robotics technologies, we’ll explore Isaac Sim, Isaac Lab, robotics simulation, digital twins, AI-driven control, and OpenUSD workflows through hands-on learning and community discussion.

What to Expect
📅 Every Wednesday from 12:30 PM to 1:30 PM PT
🎙️ Where: Discord Community Room

Together, we:

• 🤖 Explore NVIDIA Isaac Sim and Isaac Lab through hands-on learning
• 💬 Discuss robotics simulation, digital twins, Physical AI, and related technologies
• 🛠️ Build projects, experiment with workflows, and troubleshoot together
• 🎨 Share robotics projects, demos, discoveries, and learnings
• 🤝 Connect with other developers and community members building with NVIDIA robotics technologies

If You Don't Have Access to the Community Room

Attendees need to be verified members of the NVIDIA Omniverse Discord. If you just joined the Discord for the study group, follow these steps:

1. Complete the short questionnaire in #apply-for-access, located in the Welcome category.
2. Edmar or Amelia will review and approve your access.
3. Once approved, select your preferred roles in #roles so you receive the notifications most relevant to you.

Whether you're new to robotics simulation or already building with Isaac Sim and Isaac Lab, you're welcome to learn and build with us.

Join NVIDIA and Pollen Robotics for a casual coffee meetup during GTC Berlin! Come meet the team behind Microduck, connect with fellow developers and robotics enthusiasts, and chat about what's happening across robotics and Physical AI.

Whether you're curious about Microduck, exploring what's possible with Physical AI, or just looking to meet others in the community, we'd love to see you!

Everyone is welcome, whether you're attending GTC Berlin or just in the area. Grab a coffee, bring your questions, and join the conversation!

Explore what developers can build with the ideas and technology showcased at GTC Berlin. Join NVIDIA experts, model builders, and industry practitioners for conversations and demos spanning customizable open models and physical AI. Discover how these tools connect from experimenting on your desktop to powering industrial workflows and intelligent robots. Hear how builders and industry practitioners approach real use cases, discover tools and ideas for your next project, and join the conversation through live chat and audience Q&A. This stream brings the GTC experience directly to you.

Welcome to the OpenUSD Studierstube, your virtual learning space for Universal Scene Description (OpenUSD)!

Whether you’re just getting started or already working with OpenUSD, the Studierstube is a place to learn, build, share, and grow together with the community.

We started as a study group working through the official OpenUSD Learning Path to prepare for the NVIDIA OpenUSD Certification. Today, we’ve expanded to explore the broader OpenUSD ecosystem, including Digital Twins, Physical AI, and other technologies and workflows enabled by OpenUSD.

What to Expect
📅 Every Friday at 3:00 PM CEST
🎙️ Where: Discord Community Room
💬 Join the conversation: OpenUSD Studierstube Channel

Together, we:

• 📚 Work through the OpenUSD Learning Path and other resources, including OpenUSD in a Weekend
• 💬 Discuss OpenUSD, Digital Twins, Physical AI, and related technologies
• 🛠️ Build projects and experiment with new workflows
• 🎨 Share OpenUSD projects, demos, discoveries, and learnings

If You Don't Have Access to the Community Room

Attendees need to be verified members of the Omniverse Discord. If you just joined the Omniverse Discord for the study group, follow these steps:

1. Complete the short questionnaire in #apply-for-access, located in the Welcome category.
2. Edmar or Amelia will review and approve your access.
3. Once approved, select your preferred roles in #roles so you receive the notifications most relevant to you.

Come with questions, ideas, or something you’re building. All experience levels are welcome!

The OpenUSD Study Group is a dedicated space for addressing challenges and issues encountered during OpenUSD learning. As a group, we share our experiences of how we solved some of the problems and challenges we faced while learning and building OpenUSD-based solutions.Additional Resources: https://ziacreatesideas.github.io/STUDY_USD-GROUP/index.html

Welcome to the Isaac Sim & Isaac Lab Study Group, your virtual learning space for robotics simulation and development!

Led by NVIDIA Omniverse Ambassador Scott Dempsey, this interactive study group is a place to learn, build, share, and grow together with the robotics community.

Whether you’re just getting started or already building with NVIDIA robotics technologies, we’ll explore Isaac Sim, Isaac Lab, robotics simulation, digital twins, AI-driven control, and OpenUSD workflows through hands-on learning and community discussion.

What to Expect
📅 Every Wednesday from 12:30 PM to 1:30 PM PT
🎙️ Where: Discord Community Room

Together, we:

• 🤖 Explore NVIDIA Isaac Sim and Isaac Lab through hands-on learning
• 💬 Discuss robotics simulation, digital twins, Physical AI, and related technologies
• 🛠️ Build projects, experiment with workflows, and troubleshoot together
• 🎨 Share robotics projects, demos, discoveries, and learnings
• 🤝 Connect with other developers and community members building with NVIDIA robotics technologies

If You Don't Have Access to the Community Room

Attendees need to be verified members of the NVIDIA Omniverse Discord. If you just joined the Discord for the study group, follow these steps:

1. Complete the short questionnaire in #apply-for-access, located in the Welcome category.
2. Edmar or Amelia will review and approve your access.
3. Once approved, select your preferred roles in #roles so you receive the notifications most relevant to you.

Whether you're new to robotics simulation or already building with Isaac Sim and Isaac Lab, you're welcome to learn and build with us.

Serving world models for video generation requires a different balance of latency and throughput than serving large language models. In this livestream, Nebius Physical AI team members Stewart Tong and Timothy Le examine NVIDIA Cosmos 3 Super benchmarks through vLLM-Omni on NVIDIA HGX B200 and HGX H200 systems.

Their work compares four serving topologies, ranging from one eight-GPU replica to eight single-GPU replicas. See how replica count, parallelism, and concurrency affect request latency and validated video-seconds per node-hour. Learn which settings to hold constant for video generation benchmarks and how to check generated clips for file integrity and basic motion.

What You'll Learn:

• Why faster video requests can mean less output per node
• How to measure latency and throughput with video-validation checks
• When more replicas help and extra concurrency adds delay
• What NVIDIA HGX B200 and HGX H200 benchmarks show about allocating GPUs

Have questions about serving NVIDIA Cosmos 3 Super for world-model video generation? Drop them in live — the NVIDIA and Nebius teams will address them during the discussion.

Learn more about NVIDIA Cosmos 3 serving and reproduce the benchmark:
🧪 Reproduce the benchmarks → https://github.com/nebius/nebius-physical-ai
📚 Explore the NVIDIA Cosmos platform → https://github.com/nvidia/Cosmos
👥 Join the NVIDIA Omniverse community → https://discord.com/invite/nvidiaomniverse

Welcome to the OpenUSD Studierstube, your virtual learning space for Universal Scene Description (OpenUSD)!

Whether you’re just getting started or already working with OpenUSD, the Studierstube is a place to learn, build, share, and grow together with the community.

We started as a study group working through the official OpenUSD Learning Path to prepare for the NVIDIA OpenUSD Certification. Today, we’ve expanded to explore the broader OpenUSD ecosystem, including Digital Twins, Physical AI, and other technologies and workflows enabled by OpenUSD.

What to Expect
📅 Every Friday at 3:00 PM CEST
🎙️ Where: Discord Community Room
💬 Join the conversation: OpenUSD Studierstube Channel

Together, we:

• 📚 Work through the OpenUSD Learning Path and other resources, including OpenUSD in a Weekend
• 💬 Discuss OpenUSD, Digital Twins, Physical AI, and related technologies
• 🛠️ Build projects and experiment with new workflows
• 🎨 Share OpenUSD projects, demos, discoveries, and learnings

If You Don't Have Access to the Community Room

Attendees need to be verified members of the Omniverse Discord. If you just joined the Omniverse Discord for the study group, follow these steps:

1. Complete the short questionnaire in #apply-for-access, located in the Welcome category.
2. Edmar or Amelia will review and approve your access.
3. Once approved, select your preferred roles in #roles so you receive the notifications most relevant to you.

Come with questions, ideas, or something you’re building. All experience levels are welcome!

The OpenUSD Study Group is a dedicated space for addressing challenges and issues encountered during OpenUSD learning. As a group, we share our experiences of how we solved some of the problems and challenges we faced while learning and building OpenUSD-based solutions.Additional Resources: https://ziacreatesideas.github.io/STUDY_USD-GROUP/index.html

The Conference on Robot Learning (CoRL) brings together researchers, developers, and robotics teams advancing the state of the art in robot learning and Physical AI. It’s a great opportunity to explore the latest research, connect with the robotics community, and see how new approaches in foundation models, reinforcement learning, simulation, and embodied AI are moving from research into real-world robotic systems.

NVIDIA will be at CoRL 2026, and we’re working on opportunities for our Physical AI community to connect while we’re there. 

More details are coming soon, including NVIDIA activities, community meetups, and other ways to connect with NVIDIA researchers, developers, and the broader robotics community at CoRL.

Planning to attend? Let us know and share what you’re most interested in seeing or discussing at CoRL!

Join us to explore how AI is accelerating American economic growth and scientific discovery at the Ronald Reagan Building and International Trade Center.