NVIDIA at Ray Summit 2026
August 24–26 | San Francisco, CA
San Francisco Marriott Marquis
NVIDIA at Ray Summit 2026 (August 24–26, San Francisco) showcases how NVIDIA GPUs and the CUDA-X ecosystem accelerate Ray-based distributed AI workflows. NVIDIA's featured keynote covers post-training the open Nemotron model using Ray for reinforcement learning across CPUs and GPUs. Additional sessions cover vLLM inference optimization, NVIDIA GB200/GB300 topology-aware scheduling for robotics, and GPU-native data preprocessing with Ray Data and NVIDIA cuDF—which has been shown to reduce AI data processing costs by 80%
Join NVIDIA at Ray Summit 2026 to see how NVIDIA and Anyscale are partnering to make Ray run best on NVIDIA platforms. Connect with our team at the event to learn how, together, we're advancing high-performance, scalable AI and data processing.
Featured NVIDIA Conference Sessions
What Sessions is NVIDIA Presenting at Ray Summit Week?
Tuesday, August 25
3:30–3:45 p.m. PT
Wednesday, August 26
Featured NVIDIA Speakers at Ray Summit 2026
NVIDIA researchers and engineers are taking the stage at Ray Summit 2026 to share how we're pushing the boundaries of distributed AI—from post-training massive open models and serving frontier MoE architectures to accelerating data pipelines and enabling next-generation robotics. Catch their sessions to learn how NVIDIA and Ray are shaping the future of scalable AI infrastructure.
Sponsors
Frequently Asked Questions About NVIDIA at Ray Summit 2026
Ray Summit is a premier conference for the Ray community, bringing together developers and researchers to discuss scalable AI, distributed computing, and the ecosystem surrounding Ray. NVIDIA partners with Anyscale to showcase how our full-stack accelerated computing platforms, including GPUs and optimized software, enhance performance, scalability, and efficiency for Ray-based workflows.
- Explore the full conference schedule and registration details for Ray Summit 2026.
The event is scheduled for August 24–26, 2026, and will be hosted at the San Francisco Marriott Marquis in San Francisco, California.
- Review the Ray Summit event details to finalize your travel plans.
Training an open model is not only a model problem. It is a systems problem. Agentic reinforcement learning requires data generation, inference, and training to stay balanced across CPUs and GPUs. In this keynote, Bryan Catanzaro will explain how NVIDIA uses Ray to build and post-train Nemotron, what has worked, what remains hard, and how developers can apply those lessons to their own models and agentic systems. He will cover how Ray supports Nemotron's data and reinforcement learning workflows, why open models, datasets, and training recipes make it possible for developers to customize and extend the work, and where the current systems still need to improve. The goal of Nemotron is to help other people build better AI. That requires NVIDIA, Ray, and the open-source community to solve these problems together.
- View the session description and register to attend the NVIDIA Ray Summit keynote.
Ray integrates with NVIDIA technologies—such as cuDF and the broader CUDA-X ecosystem—to offload data preprocessing and training tasks directly to the GPU. This integration enables developers to create GPU-native, end-to-end data pipelines that significantly increase throughput while reducing compute costs.
- Read about how Anyscale optimizes multimodal AI data processing using NVIDIA RTX acceleration.
Registration for all sessions, including the NVIDIA keynote and co-located events, is managed directly through the official Ray Summit event portal.
- Secure your spot and view the full agenda at the Ray Summit 2026 registration page.
By integrating Ray Data with NVIDIA cuDF, organizations can perform multimodal data preprocessing directly on the GPU, avoiding the overhead of transferring data between CPU and GPU. This optimization has been shown to cut data processing costs significantly, such as the 80% reduction achieved by Anyscale using NVIDIA RTX PRO.
- Read the case study on lowering AI infrastructure costs through GPU-accelerated pipelines.
Yes, the summit features a session on Ray's topology-aware scheduling on NVIDIA GB200 and GB300 systems, which specifically addresses the unique performance and orchestration requirements for physical robotics and GR00T models.
- Check the Ray Summit session catalog for technical talks related to robotics and AI orchestration.
The Ray Summit page includes a dedicated speaker section featuring NVIDIA experts like Bryan Catanzaro, Siyuan Fu, and others, along with links to their respective session topics.
- View the list of featured NVIDIA speakers and their scheduled technical sessions.
- Watch the video on scaling multimodal data curation to see architectural patterns for petabyte-scale pipelines.
Topology-aware scheduling enables the Ray scheduler to recognize the underlying physical layout of the GPU cluster, such as memory and interconnect pathways. This awareness minimizes data travel times and optimizes workload placement on architectures like the NVIDIA GB200 and GB300, leading to significantly better performance.
- Explore the Ray Summit technical sessions for discussions on cluster resource management.
Yes, you can subscribe to the NVIDIA newsletter through the event page to stay informed about the latest developments and future event notifications.
- Visit the official NVIDIA event landing page to subscribe for updates.
NVIDIA Resources for Ray Summit 2026 Attendees
Sign up to receive the latest news from NVIDIA.