NVIDIA at SGLang Summit

NVIDIA at SGLang Summit

November 12–13 | Fort Mason, San Francisco

Stay ahead of the latest open source inference breakthroughs and plug into a thriving community driving the field forward.

Register Now

    Be part of the first-ever SGLang Summit, where experts, practitioners, and builders will come together for hands-on training and deep-dive sessions on how to get the most out of SGLang on NVIDIA. Learn what makes it perform best and how to deploy it with confidence.

    Meet NVIDIA Leaders Driving AI Innovation

    NVIDIA Keynote on Open Models

    Explore open frontier models in a fireside and panel chat with Joey Conway, NVIDIA senior director of product management, deep learning software.

    View Details

    NVIDIA Keynote on Vera Rubin

    Join Ujval Kapasi, NVIDIA vice president of AI and HPC frameworks and libraries, to explore the latest in AI hardware, including NVIDIA Vera Rubin.

    View Details

    Schedule at a Glance

    All times are in Pacific. Main-stage times are tentative and likely to change.

    11:45 a.m.–12:30 p.m.

    Frontier Models Fireside and Panel Chat — Main Stage

    Joey Conway

    1:30–1:50 p.m.

    SGLang for Agents on NVIDIA Vera Rubin — Main Stage

    Ujval Kapasi

    1:30–3:30 p.m.

    NVIDIA Local AI Coding Lab — Prefill Room

    Jerry (LMSYS) delivering; Mackenzie Kosseim, Lakshmi Ramesh, Ankur Singh, Arjun Aiyer supporting

    4:30–4:50 p.m.

    Getting to Speed-of-Light Performance on NVIDIA at Speed-of-Light Velocity with Agents — Main Stage

    Luis Ceze

    5:10–5:25 p.m.

    Bringing Up and Optimizing SGLang on NVIDIA Vera Rubin — RadixAttention Room

    Lee Nau, Akhil Goel

    10 a.m.–12 p.m.

    Hands-On Reinforcement Learning With SGLang and Miles — Prefill Room

    Hunter Carlisle, Kaixi Hou

    1:00–1:20 p.m.

    Taking SGLang to Enterprise Production With NVIDIA NIM — RadixAttention Room

    Arun Venkatesan, Arthur Cheng, Shreyas Misra

    1:20–1:30 p.m.

    Evaluation (SWE-Serve) — RadixAttention Room

    Jiantao Jiao

    3:30–3:45 p.m.

    Closing the Loop: From Agentic Evaluation at Scale to Actionable Model Insights

    Christos Kozyrakis

    3:30–3:50 p.m.

    Closing the Loop: From Agentic Evaluation at Scale to Actionable Model Insights — RadixAttention Room

    Meriem Boubdir

    5:10 p.m. – 5:25 p.m

    Day Zero on Vera Rubin: SGLang, Miles and FlashInfer — HiCache Room

    Jingfan Sun, Yang Xu

    5:10–5:30 p.m.

    Dynamo and SGLang: Scaling Agentic Inference and RL Rollouts — RadixAttention Room

    Ria Jain, Biswa Ranjan Panda, Hyunjae Woo

    Hear About the Latest Breakthroughs From Our Experts

    Joey Conway

    Sr Director of Product Management, Deep Learning Software

    Ujval Kapasi

    VP, AI and HPC Frameworks and Libraries

    Jiantao Jiao

    Director of AI and Distinguished Scientist

    Luis Ceze

    VP, AI Systems Software

    Akhil Goel

    Software Engineer

    Arthur Cheng

    Principal Engineer, NVIDIA NIM™

    Arun Venkatesan

    Group Product Manager, AI Inference

    Biswa Ranjan Panda

    Senior Software Engineer

    Hunter Carlisle

    Principal, Open Source Reinforcement Learning Frameworks

    Hyunjae Woo

    Senior Systems Software Engineer, NVIDIA Dynamo

    Jingfan Sun

    Software Engineering Manager, CUDA™ Kernels and Libraries

    Kaixi Hou

    Senior Deep Learning Software Engineer

    Lee Nau

    Senior Deep Learning Software Engineer, Inference

    Meriem Boubdir

    Senior Deep Learning Engineer, LLM Accuracy Evaluation

    Ria Jain

    Deep Learning Algorithms Engineer

    Shreyas Misra

    Senior Deep Learning Algorithms Engineer

    Yang Xu

    Senior Manager, Deep Learning Software

    SGLang Summit NVIDIA and Modal Meetup

    The night before the summit, join NVIDIA, SGLang, and Modal for an evening of lightning talks, inference roundtables, a live survey and raffle, and food, drinks, and conversations.

    Whether you’re in town early or just want to connect with the inference community in SF, this is your chance to grab a drink, meet the engineers and teams behind the tools you use, and talk shop about what’s working, what’s not, and what’s next in inference.

    Request to Attend

    Special Events

    DigitalOcean, NVIDIA, SGLang, and Major League Hacking Meetup

    Learn how teams build, test, and improve AI agents on one platform, from inference to deployment. See how they run coding agents safely, evaluate performance on real workloads, choose the right open model for each task, and use production insights to improve.

    View Details

    SGLang Summit NVIDIA Meetup

    Join for demos, drinks, and developer conversations the night before the first SGLang Summit.

    Meet the engineers behind your tools and connect with San Francisco’s open models and inference community to discuss what’s working and what’s next.

    View Details

    Explore Our Work With SGLang

    Get a glimpse into what you can expect by exploring past NVIDIA On Demand content. Behind these efforts are dedicated engineers ensuring the code evolves to meet the needs of both researchers and developers.

    Baizhou Zhang, Qiaolin Yu, Yueming Yuan, Radixark.ai

    High-Performance LLM Serving and Training With SGLang

    This training lab shows you how to optimize and scale large language model (LLM) workflows with SGLang. We walk through practical performance tuning using the SGL-Cookbook, dive into profiling and bottleneck analysis, and demonstrate its deep integration into reinforcement learning (RL) training frameworks.

    Watch Now
    Trevor Morris, SGLang Committer, NVIDIA

    Driving AI Inference Performance: Q&A

    In this Q&A, Trevor Morris, senior deep learning software engineer at NVIDIA, discusses his role as an SGLang committer, recent performance optimizations, and how these contributions are helping to strengthen the open source ecosystem around LLMs.

    Read More
    Baizhou Zhang, Core Developer, SGLang

    Inference OSS Ecosystem Featuring SGLang

    This session shows how SGLang and NVIDIA Dynamo enable production DeepSeek serving on NVIDIA GB200 NVL72. Learn how disaggregated prefill/decode, large-scale parallelism, multi-token prediction (MTP) scheduling, and compute–communication overlap deliver lower latency and multi-x speedups for mixture-of-experts (MoE) workloads.

    Watch Now

    FAQ

    The SGLang Summit is the first-ever event bringing together experts, practitioners, and builders for hands-on training and deep-dive sessions on getting the most out of SGLang on NVIDIA. It takes place on November 12–13 at Fort Mason in San Francisco.

    The summit is designed for developers, programmers, IT professionals, AI researchers, and technical decision-makers who want to stay ahead of the latest open source inference breakthroughs.

    Featured speakers include Joey Conway (Senior Director of Product Management, Deep Learning Software), Ujval Kapasi (VP, AI and HPC Frameworks and Libraries), Jiantao Jiao (Director of AI and Distinguished Scientist), and Luis Ceze (VP, AI Systems Software).

    SGLang integrates with NVIDIA Dynamo to deliver production-ready serving on advanced hardware like the GB200 NVL72, supporting multi-token prediction scheduling and disaggregated prefill/decode architectures for extreme speedups.

    By leveraging large-scale parallelism, disaggregated prefill/decode, and multi-token prediction scheduling, SGLang delivers lower latency and multi-x speedups for MoE workloads on NVIDIA systems.

    Check out SGLang technical documentation on NVIDIA Docs.

    Company Information
    News and Events
    Popular Links