November 12 - 13, 2026. Fort Mason, San Francisco

SGLANGSUMMIT2026

Open Foundations for the Next Era of AI.

TRUSTED BY INDUSTRY LEADERS

Expo Pass

Two days of technical talks, hands-on workshops and connections with the frontier AI community.

Technical
Breakouts

AI in Production

Cloudflare shares lessons from large-scale deployment. LinkedIn explains how it serves a small model for ads ranking at over 10K queries per second with scoring latency under 80 ms. Other sessions cover multimodal data pipelines and private inference.

SGLang & Miles

Get the latest on the core engine and Miles from the teams building them. Hear how Meta brings PyTorch to SGLang and how Alibaba is extending HiCache to support long-running agents.

Agentic Engineering

Perplexity shares a minimalist trainer designed for trillion-parameter models, built with help from coding agents. Modal presents early results from using agents to develop and manage SGLang inference services.

Training & Post-Training

Training-system sessions feature Thinking Machines Lab and Meta’s work on TorchTitan with Miles. Decagon shares how RL improves instruction following and reduces hallucinations in customer-facing agents.

AI Hardware

NVIDIA engineers explain how they brought SGLang to Vera Rubin and optimized its performance. Google shares the development and deployment of SGLang-TorchTPU for production inference at global scale.

Real-Time AI & Media

SGLang Omni handles conversations where users can interrupt and talk while the model is speaking. HeyGen explains how GPU optimization and scheduling cut serving costs, while MiniMax covers AI workflows for film production.

AI for Science

Harmonic shows how machine-checked proofs can catch errors in AI-generated software and hardware. Periodic Labs discusses the methods and infrastructure that let laboratories learn from each experiment.

Robotics

Generalist shares the research behind GEN-1.5 and its ability to pick up new skills on the fly. RLinf covers reinforcement learning across simulation and real robots, drawing on deployment across 96 robots.

Main Stage
Livestream

Watch Main Stage talks and panels live on screens at the venue, with discussions on frontier models, AI hardware, recursive self-improvement and agents.

Hands-On
Workshops

Train a model with reinforcement learning using SGLang and Miles in NVIDIA’s workshop. Compare AMD and NVIDIA GPU performance with Crusoe, and learn to speed up agent tasks with Nebius. AMD and Modal also host hands-on sessions.

Meet the
Community

Meet AI engineers, researchers and agent builders at sponsor booths and in the lounges. Visit the GPU Expo and see Tesla’s Optimus up close. Enjoy complimentary breakfast and lunch, plus Kyoto Matcha, Hedge Coffee and tanghulu.

Speakers

LIP-BU
TAN

CEO

LILIAN
WENG

Member of Technical Staff

SOUMITH
CHINTALA

CTO

ARAVIND
SRINIVAS

CEO & Co-Founder

VAMSI
BOPPANA

Senior Vice President, Artificial Intelligence

BILL
JIA

VP of Engineering - Core ML/AI

RUOMING
PANG

Member of Technical Staff

UJVAL
KAPASI

VP, AI & HPC Frameworks and Libraries

LUIS
CEZE

VP, AI Systems Software, Compilers

COLIN
BRACE

Vice President of Technology and Software

LIAM
FEDUS

Co-Founder & Co-CEO

ERIC
ZELIKMAN

CEO, Co-Founder & MTS

MARK
SAROUFIM

Co-Founder & MTS

YUANDONG
TIAN

Co-Founder

LINDEN
LI

Co-Founder & Chief Architect

DYLAN
PATEL

Founder & CEO

NANCY
ZHENG

VP of Engineering in Service Platform, Disagg

JOEY
CONWAY

Senior Director, Product Management

BEHROOZ
GHORBANI

Reinforcement Learning Research Lead

YUN
JIN

Head of AI Infra

RICHARD
GONG

Head of Inference

KEVIN
FLANSBURG

Senior Manager, Engineering

SYONA
SARMA

VP of Hardware Engineering & Operations

NITIN
PERUMBETI

CTO

PARTH A.
PATEL

Co-Founder

EDWARD
HU

Head of AI Modeling

ALEX
SHAW

Member of Technical Staff

LIANMIN
ZHENG

AI Research Scientist

YING
SHENG

Co-Founder & CEO

BANGHUA
ZHU

Co-Founder & CTO

View All Speakers

Agenda

Frontier Infrastructure, Silicon, Models, and Applications, on One Stage

SGLang Summit brings together the engineers, researchers, builders, and leaders behind the open AI stack.

Day 1 Thu, Nov 12

9:00AM
Full Pass check-in opens

10:00AM
10:00 - 10:20 AMMain Stage · Full Pass Only
SGLang
Lianmin Zheng AI Research Scientist, MetaYing Sheng Co-Founder & CEO, RadixArk
10:25 - 10:35 AMMain Stage · Full Pass Only
SGLang Ecosystem & Miles
Banghua Zhu Co-Founder & CTO, RadixArk
10:30AM
10:35 - 10:55 AMMain Stage · Full Pass Only
Fireside Chat with Vamsi Boppana
Vamsi Boppana Senior Vice President, Artificial Intelligence, AMD
11:00AM
11:00 - 11:30 AMMain Stage · Full Pass Only
Fireside Chat with Lilian Weng
Lilian Weng Member of Technical Staff, OpenAI
11:30AM
11:45 - 12:30 PMMain Stage · Full Pass Only
Panel: Frontier Models
Joey Conway Senior Director, Product Management, NVIDIARuoming Pang Member of Technical Staff, OpenAIBehrooz Ghorbani Reinforcement Learning Research Lead, Reflection AI
12:30 - 1:30PM
Lunch
1:00PM
Expo Pass check-in opens

1:30PM
1:30 - 1:50 PMMain Stage · Full Pass Only
SGLang for Agents on NVIDIA Vera Rubin
Ujval Kapasi VP, AI & HPC Frameworks and Libraries, NVIDIA
1:50 - 2:20 PMMain Stage · Full Pass Only
Fireside Chat with Soumith Chintala
Soumith Chintala CTO, Thinking Machines Lab
1:30 - 2:00 PMRadixAttention Room · Full & Expo Pass
SGLang for the Agentic Era: Powering Coding Agents and Automated Optimization
Emad Barsoum CVP, AI, AMD
1:30 - 2:00 PMHiCache Room · Full & Expo Pass
Fast Models, Diverse Workloads: Optimizing SGLang for Intel GPUs and Xeon CPUs
Mingfei Ma Senior Deep Learning Software Architect, Intel
1:30 - 3:00 PMDecode Room · Full & Expo Pass
Nebius Advanced GPU Inference Optimization with Nebius: Hardware, Compression, Kernels, and Serving
Ryan Hanrui Wang SVP, Applied AI, NebiusSamir Khaki Principal AI Engineer, NebiusWei-Chen Wang VP of AI Inference, NebiusRoy Cheng Nebius
1:30 - 3:30 PMPrefill Room · Full & Expo Pass
LMSYS Local AI Coding Lab
Jerry Zhou Member of Technical Staff, Microsoft AILakshmi Ramesh Senior Software Engineer, NVIDIAMackenzie Kosseim Senior Product Manager, NVIDIAAnkur Singh Senior Systems Software Engineer, NVIDIAArjun Aiyer Developer Marketing Manager, NVIDIA
2:00PM
2:20 - 3:05 PMMain Stage · Full Pass Only
Panel: Recursive Self-Improvement
Yuandong Tian Co-Founder, RecursiveMark Saroufim Co-Founder & MTS, Core Automation
2:00 - 2:20 PMRadixAttention Room · Full & Expo Pass
The Gateway to the Agentic Enterprise
Amy Yuan Sr. Director, Engineering Systems + TPM, SnowflakeYusuf Ozuysal Technical Assistant, Office of the CEO, Snowflake
2:20 - 2:35 PMRadixAttention Room · Full & Expo Pass
Scaling Frontier Model Training at Osmosis
Andy Lyu CTO, Osmosis AI
2:00 - 2:20 PMHiCache Room · Full & Expo Pass
Prefill Here, Decode There: Disaggregating Inference Across Datacenters and Hardware Classes
Matt Murphy Member of Technical Staff, Fireworks AI
2:20 - 2:40 PMHiCache Room · Full & Expo Pass
DFlash and Beyond: How Inco Runs the Fastest Open-Model Endpoints
Zhijian Liu Co-Founder & Chief Scientist, Inco AI
2:30PM
2:45 - 3:00 PMRadixAttention Room · Full & Expo Pass
High-Efficiency, Production-Ready Inference
Liangsheng Yin Member of Technical Staff, RadixArk
2:50 - 3:05 PMHiCache Room · Full & Expo Pass
Aristotle: Formal Verification, Today
Tudor Achim CEO, Harmonic
3:00PM
3:20 - 3:40 PMMain Stage · Full Pass Only
Building What's Next: How Trainium and SGLang Are Shaping the Future of AI Infrastructure
Colin Brace Vice President of Technology and Software, AWS
3:00 - 3:15 PMRadixAttention Room · Full & Expo Pass
Agents That Write Megakernels: Multi-Day Campaigns for GPU Performance in SGLang
Chengze Fan Software Engineer, MetaOguz Ulgen Software Engineer, Meta Superintelligence Labs
3:15 - 3:30 PMRadixAttention Room · Full & Expo Pass
Scaling SGLang on AMD Hardware with ROCm and Open Source
Hai Xiao Fellow, ML and AI Software Architect, AMD
3:05 - 3:20 PMHiCache Room · Full & Expo Pass
Training Reliable Customer Service Agents with Large-Scale RL
Cyrus Asgari Research Tech Lead, Decagon
3:20 - 3:35 PMHiCache Room · Full & Expo Pass
Tinker API Training with Miles
Chuck Tang Research Engineer, Scaling RL Compute, Trajectory
3:10 - 3:25 PMDecode Room · Full & Expo Pass
Training One Policy Across Many Agent Harnesses: Agentic RL Post-Training with SGLang, Miles and Harbor
Tianqi Liu Member of Technical Staff, HarkJackie Yuan Member of Technical Staff, Hark
3:25 - 3:40 PMDecode Room · Full & Expo Pass
The Path to Inference Efficiency
Christos Kozyrakis Stanford University & NVIDIA Research
3:05 - 3:20 PMOpen Stage · Full & Expo Pass
Beyond Multimodal: Building Full-Modal Intelligence with Open Weights
Ryan Lee Head of Technical Ecosystem, MiniMax
3:30PM
3:40 - 4:00 PMMain Stage · Full Pass Only
500 Tokens per Second Is the New Slow
Nancy Zheng VP of Engineering in Service Platform, Disagg, Cerebras
3:30 - 3:50 PMRadixAttention Room · Full & Expo Pass
Memory Accelerated Prefix Serving for HiCache Tiering
Vasanthi Jagatha Sr. Product Manager, AI Memory Innovations, SamsungSomnath Roy Sr. Principal Engineer, AI Memory Innovations, Samsung
3:35 - 3:50 PMHiCache Room · Full & Expo Pass
Building Labs That Learn
Byron Hsu Member of Technical Staff, Periodic Labs
3:40 - 3:55 PMDecode Room · Full & Expo Pass
From Breakthroughs to Builders: The Dawn of Physical AI
Tao Chen CEO, Dexmate
3:50 - 5:20 PMPrefill Room · Full & Expo Pass
Crusoe Serve, Measure, Tune: SGLang on Crusoe Managed Kubernetes
Rohit Kalmankar Senior Staff Solutions Engineer, CrusoeMartin Cala Staff Solutions Engineer, Crusoe
4:00PM
4:00 - 4:30 PMMain Stage · Full Pass Only
Fireside Chat with Lip-Bu Tan
Lip-Bu Tan CEO, Intel
4:00 - 4:20 PMRadixAttention Room · Full & Expo Pass
Same SGLang, New Neuron: Inference and RL Rollouts on Trainium with TorchNeuron
Pinak Panigrahi Annapurna ML, AWSAlex Nails Member of Technical Staff, RadixArk
4:20 - 4:40 PMRadixAttention Room · Full & Expo Pass
AgentX × SGLang: SGLang on Real Agent Workloads
Wenyao Gao Member of Technical Staff, SemiAnalysis
4:00 - 4:20 PMHiCache Room · Full & Expo Pass
From Fast Tokens to Fast Agents
Ryan Hanrui Wang SVP, Applied AI, Nebius
4:20 - 4:40 PMHiCache Room · Full & Expo Pass
Efficient Long-Context Inference with SGLang: Lessons from Xiaomi MiMo
Chen Zhang Head of the MiMo Inference Team, XiaomiYingchun Lai Senior Inference Engineer, Xiaomi
4:00 - 4:15 PMDecode Room · Full & Expo Pass
One Model, Many Chips. The Challenges of Inference on Heterogeneous Compute
Piotr Mazurek Member of Technical Staff, Inference, Liquid AI
4:15 - 4:30 PMDecode Room · Full & Expo Pass
Boreal: Towards Self-Improving Video Generation
Ledell Wu Co-Founder & Chief Scientist, Creatify AI
4:30PM
4:35 - 4:55 PMMain Stage · Full Pass Only
Getting to Speed-of-Light Performance on NVIDIA at Speed-of-Light Velocity with Agents
Luis Ceze VP, AI Systems Software, Compilers, NVIDIA
4:50 - 5:10 PMRadixAttention Room · Full & Expo Pass
Bringing Up and Optimizing SGLang on NVIDIA Vera Rubin
Lee Nau Senior Deep Learning Software Engineer, Inference, NVIDIAAkhil Goel Software Engineer, NVIDIA
4:50 - 5:15 PMHiCache Room · Full & Expo Pass
Every Spare Gigabyte: A Fleet-Wide Elastic Mooncake Cache
Kastan Day Machine Learning Engineer, Cloudflare
4:30 - 4:45 PMDecode Room · Full & Expo Pass
Agents Need Safe Spaces Too
Chang She CEO & Co-Founder, LanceDB
4:45 - 5:00 PMDecode Room · Full & Expo Pass
Science at Inference Speed: Fast, Open Model Inference for Science Agents
Prasanna Balaprakash Co-Founder & CEO, PrimaLabs
5:00PM
5:10 - 5:30 PMRadixAttention Room · Full & Expo Pass
Dynamo and SGLang: Scaling Agentic Inference and RL Rollouts
Ria Jain Deep Learning Algorithms Engineer, NVIDIABiswa Ranjan Panda Engineer, Dynamo, NVIDIAHyunjae Woo Software Engineer, NVIDIA
5:15 - 5:30 PMHiCache Room · Full & Expo Pass
From HiCache to Agentic KV Infrastructure: Evolving SGLang’s Hierarchical Memory and Unified Caching at Production Scale
Wei Zhang Distinguished Engineer & Senior Director, Alibaba
5:00 - 5:15 PMDecode Room · Full & Expo Pass
Scaling LLM Inference in Production: Lessons Learned with AIBrix Gateway
Varun Gupta Staff Software Engineer, ByteDance
5:15 - 5:35 PMDecode Room · Full & Expo Pass
Train for the SLA: Post-Training Models That Meet Production Latency and Cost Targets
Tianyou Li Member of Technical Staff, NeuroSpark AIThomas Ning Founder, NeuroSpark AI
5:15 - 5:30 PMOpen Stage · Full & Expo Pass
SGLang: Qualcomm Dragonfly's Native Backend Contribution
Parmeet Kohli Director, Product Management, QualcommAnkit Arora Senior Staff Engineer, Data Center AI, Qualcomm

Day 2 Fri, Nov 13

9:00AM
Full Pass and Expo Pass check-in opens

10:00AM
10:00 - 10:30 AMMain Stage · Full Pass Only
AI for Humans
Eric Zelikman CEO, Co-Founder & MTS, humans&
10:00 - 10:15 AMRadixAttention Room · Full & Expo Pass
Unified Runtime for Full-Duplex Interactive Intelligence with SGLang Omni
Chenyang Zhao Founding Member of Technical Staff, RadixArkJintao Qu SGLang Omni Team
10:15 - 10:30 AMRadixAttention Room · Full & Expo Pass
Building Real-Time Multimodal Agents: Experience, Intelligence, and Economics
Mu Li Co-Founder, CTO & President, Boson AI
10:00 - 10:15 AMHiCache Room · Full & Expo Pass
Two Mental Models About Inference that I Find Useful and I Hope You Do Too
Horace He Member of Technical Staff, Thinking Machines
10:15 - 10:30 AMHiCache Room · Full & Expo Pass
Train on One Chip, Roll Out on Any: A Field Guide to RL on Heterogeneous Hardware
Yağız Çalık Founding Member of Technical Staff (Post-Training), ai&
10:00 - 10:15 AMDecode Room · Full & Expo Pass
Superhuman Evals for Superhuman Intelligence
Xiangyi Li Founder & CEO, BenchFlow
10:15 - 10:30 AMDecode Room · Full & Expo Pass
The Last Mile of AI Infra: Bring Agents to Millions of People Where They Already Are
Daniel Tian CEO & Co-Founder, Photon
10:00 - 12:00 PMPrefill Room · Full & Expo Pass
NVIDIA Hands-On RL with SGLang and Miles
Hunter Carlisle Principal, Open-Source RL Frameworks, NVIDIAKaixi Hou Senior Deep Learning Software Engineer, NVIDIA
10:30AM
10:30 - 11:00 AMMain Stage · Full Pass Only
Fireside Chat with Liam Fedus
Liam Fedus Co-Founder & Co-CEO, Periodic Labs
10:30 - 10:45 AMRadixAttention Room · Full & Expo Pass
Every GPU-Second Counts: Cost-Efficient Model Serving at HeyGen
Rui Zhang Head of Core Engineering, HeyGen
10:45 - 11:00 AMRadixAttention Room · Full & Expo Pass
Agents All the Way Down: Voice AI in an Agentic World
Rissa Cao Co-Founder & CEO, Fish Audio
10:30 - 10:45 AMHiCache Room · Full & Expo Pass
Ultrafast Inference on SGL
Tri Dao Model Performance Engineer, BasetenAlexander Korte Model Performance Engineer, Baseten
10:45 - 11:00 AMHiCache Room · Full & Expo Pass
Designing Inference Systems in the Age of Research
Francis Zhang Co-Founder, Core Automation
10:30 - 10:45 AMDecode Room · Full & Expo Pass
Agent Inference at Hardware Speed: Co-Designing GPU Kernels and Compilers
Zhihao Jia CTO, LithosAI
10:45 - 11:00 AMDecode Room · Full & Expo Pass
Recent Advances in HiCache
Zhiqiang Xie Research Scientist, Meta
11:00AM
11:00 - 11:30 AMMain Stage · Full Pass Only
Production Serving at Meta with SGLang
Lianmin Zheng AI Research Scientist, Meta
11:10 - 11:40 AMRadixAttention Room · Full & Expo Pass
Global Scale Production Inference on TPUs with SGLang-TorchTPU
Qi Zhou Principal Software Engineer, GoogleWangyuan Zhang Software Engineer, Google
11:10 - 11:25 AMHiCache Room · Full & Expo Pass
Miles: Production-Level Post-Training
Zhichen Zeng Member of Technical Staff, RadixArk
11:25 - 11:40 AMHiCache Room · Full & Expo Pass
Miles-Diffusion
Kangrui Du Member of Technical Staff, RadixArk
11:10 - 11:25 AMDecode Room · Full & Expo Pass
Build a Task-Specific RLM with SIE
Daniel Svonava Founder & CEO, Superlinked
11:25 - 11:40 AMDecode Room · Full & Expo Pass
OpenRSI × SGLang-Omni: Scaling RSI for AI Infra
Chenyang Zhao Founding Member of Technical Staff, RadixArk / OpenRSIZhuofeng Li Student Researcher & OpenRSI Co-Lead, Google DeepMind
11:30AM
11:40 - 12:00 PMRadixAttention Room · Full & Expo Pass
MOP: A Self-Learning Agent for SGLang Recipe Optimization
Balaji Varadarajan Staff Engineer/Architect, LLM Inference, DigitalOcean
11:40 - 12:00 PMHiCache Room · Full & Expo Pass
Making SGLang More Resource-Efficient: Explorations from Tencent TACO
Xibo Sun Senior AI Infrastructure Engineer, Tencent TACO
11:40 - 11:55 AMDecode Room · Full & Expo Pass
Serving Reflection Beam in SGLang
Kaixing Wu Member of Technical Staff, Reflection AI
11:30 - 1:00PM
Lunch
1:00PM
1:00 - 1:30 PMMain Stage · Full Pass Only
Resilient, Elastic Infrastructure for Continual Learning
Linden Li Co-Founder & Chief Architect, Applied Compute
1:00 - 1:20 PMRadixAttention Room · Full & Expo Pass
Taking SGLang to Enterprise Production with NVIDIA NIM
Arthur Cheng Principal Engineer, NIM, NVIDIAArun Venkatesan Group Product Manager, NVIDIAShreyas Misra Senior Deep Learning Algorithms Engineer, NVIDIA
1:20 - 1:40 PMRadixAttention Room · Full & Expo Pass
Closing the Loop: From Agentic Evaluation at Scale to Actionable Model Insights
Jiantao Jiao Director of Research & Distinguished Scientist, NVIDIAMeriem Boubdir Senior Deep Learning Engineer, LLM Accuracy Evaluation, NVIDIA
1:00 - 1:20 PMHiCache Room · Full & Expo Pass
autoinference
Charles Frye Member of Technical Staff, Modal
1:20 - 1:40 PMHiCache Room · Full & Expo Pass
Challenges and Learnings of Building a Post-Training Platform
Yiwei Song Senior Director, CrusoeDaniel Shen ML Engineer, Crusoe
1:00 - 1:15 PMDecode Room · Full & Expo Pass
Teaching Robots in the Moment: The Story of GEN-1.5
Felix Wang Research Scientist, Generalist
1:15 - 1:30 PMDecode Room · Full & Expo Pass
RL Is Everything, Everywhere, All at Once
Zhanghao Wu Co-Founder, SkyPilot
1:00 - 2:30 PMPrefill Room · Full & Expo Pass
AMD Scalable Agentic RL on MI355X with Miles
Liz Li Principal AI Engineer, AMDSree Rohith Pulipaka Sr. Software Engineer, AI, AMD
1:30PM
1:30 - 2:00 PMMain Stage · Full Pass Only
Building the Next Computer
Aravind Srinivas CEO & Co-Founder, Perplexity
1:50 - 2:50 PMRadixAttention Room · Full & Expo Pass
From Open Model to Live Stories
Mick Qian Member of Technical Staff, RadixArkLeanna Ren VP, Global Marketing & Partnerships, MiniMaxVictor Su-Ortiz Global GTM & Developer Relations, MiniMaxKenneth Loebenberg Creative Engineer, falBernie Su Co-Founder, Pickford
1:50 - 2:20 PMHiCache Room · Full & Expo Pass
TorchTitan x SGLang: One Model from Training to Rollout / How Meta Integrates with SGLang from OSS Model Optimization to Our PyTorch-Native Stack
Tianyu Liu PyTorch Engineering, MetaJana van Greunen Director of PyTorch Engineering, Meta
1:30 - 1:45 PMDecode Room · Full & Expo Pass
Building the Grid for AI Compute
Jasper Zhang Founder & CEO, Hyperbolic
1:45 - 2:00 PMDecode Room · Full & Expo Pass
Private Inference at 200B Tokens/Day: Production with dstack and SGLang
Hang Yin Co-Founder & CTO, Phala
2:00PM
2:00 - 2:30 PMMain Stage · Full Pass Only
Fireside Chat with Bill Jia
Bill Jia VP of Engineering - Core ML/AI, Google
2:20 - 2:35 PMHiCache Room · Full & Expo Pass
Production Blackwell-Native MXFP8 & NVFP4 RL with SGLang and Miles
Ziang Li Founding Member of Technical Staff, humans&
2:10 - 2:25 PMDecode Room · Full & Expo Pass
Don’t Over-Buy HBM, Tier It! (Why XPUs Need External Memory Hierarchy)
Nikita Lazarev Co-Founder & CTO, Netpreme
2:25 - 2:40 PMDecode Room · Full & Expo Pass
Running Trillion Parameter LLMs on Consumer Hardware with SSD
Zili Meng Co-Founder, WiCi AI
2:30PM
2:45 - 3:30 PMMain Stage · Full Pass Only
AI Compute Panel
Yun Jin Head of AI Infra, Fireworks AIRichard Gong Head of Inference, ModalKevin Flansburg Senior Manager, Engineering, CloudflareSyona Sarma VP of Hardware Engineering & Operations, DigitalOceanNitin Perumbeti CTO, Crusoe
2:35 - 2:50 PMHiCache Room · Full & Expo Pass
Speculative Decoding in Production: Choosing Between EAGLE and DFlash
Nicholas Liu Inference Lead, Decagon
2:40 - 2:55 PMDecode Room · Full & Expo Pass
Where Will Margin Live? Token, GPU, or Model?
Ethan Anderson COO, SF Compute
2:50 - 3:50 PMPrefill Room · Full & Expo Pass
Modal Inference from First Principles
Gilford Ting Member of Technical Staff, Modal
2:30 - 2:45 PMOpen Stage · Full & Expo Pass
Getting More from Local Hardware
Yannik Marek Creator of ComfyUI
3:00PM
3:00 - 3:15 PMRadixAttention Room · Full & Expo Pass
Beyond the Fast SGLang Engine: Building LinkedIn's AI Serving Infrastructure
Qing Lan Sr. Manager, AI Infra, LinkedIn
3:15 - 3:30 PMRadixAttention Room · Full & Expo Pass
Ads Ranking with LLMs @ LinkedIn
Xianxing Zhang Distinguished Engineer, LinkedIn
3:00 - 3:30 PMHiCache Room · Full & Expo Pass
Beyond Defaults: How Baseten Builds Production-Grade Speed and Stability on Top of SGLang
Pankaj Gupta Co-Founder, Baseten
3:00 - 3:15 PMDecode Room · Full & Expo Pass
Building a Unified Ecosystem for Optimized Kernels
Adarsh Shirawalmath Machine Learning Engineer, Hugging Face
3:15 - 3:30 PMDecode Room · Full & Expo Pass
Bringing SGLang to Spyre in a Week
Antoni Viros i Martin Staff Research Scientist, IBM Research
3:30PM
3:45 - 4:00 PMMain Stage · Full Pass Only
Teaching Frontier Models to Do Real Work
Edward Hu Head of AI Modeling, Mercor
3:30 - 3:45 PMRadixAttention Room · Full & Expo Pass
Engineering a Minimalist LLM Trainer with Coding Agents
Lequn Chen Member of Technical Staff, Perplexity
3:45 - 4:00 PMRadixAttention Room · Full & Expo Pass
Scaling Multimodal Data Curation with SGLang and KubeRay at Hark
Kai-Hsun Chen Member of Technical Staff, HarkCindy Zhang Software Engineer, Hark
3:30 - 3:45 PMHiCache Room · Full & Expo Pass
SGLang-TorchTPU: PyTorch-Native LLM Serving on TPU
Haolin Fu Member of Technical Staff, RadixArk
3:45 - 4:00 PMHiCache Room · Full & Expo Pass
MoE Optimization on SGLang-Trainium
Dong Li Chief Scientist, Yotta Labs
3:30 - 3:45 PMDecode Room · Full & Expo Pass
From Chips to Clusters: Open-Source Simulation for LLM Serving at Scale
Zhenhua Zhu Assistant Professor, Tsinghua Shenzhen International Graduate School
3:45 - 4:00 PMDecode Room · Full & Expo Pass
Serving Small Language Models with SGLang for Malicious Traffic Detection
Zhemin Su Software Engineer, Palo Alto NetworksMatthew Schulz Staff Software Engineer, Palo Alto Networks
4:00PM
4:00 - 4:15 PMMain Stage · Full Pass Only
Software Engineering Became Machine Learning
Alex Shaw Member of Technical Staff, Laude Institute
4:15 - 4:30 PMMain Stage · Full Pass Only
The State of RL Environments and Data
Parth A. Patel Co-Founder, HUD
4:10 - 4:25 PMRadixAttention Room · Full & Expo Pass
Mooncake: KVCache-Centric Infrastructure for LLM Serving
Shangming Cai Mooncake/SGLang Contributor, RadixArk
4:25 - 4:40 PMRadixAttention Room · Full & Expo Pass
SGLang Serves the Model. llm-d Orchestrates the Fleet.
Akshay Ram Group Product Manager, Google
4:10 - 4:25 PMHiCache Room · Full & Expo Pass
From Python to Rust: Reengineering SGLang Server
Rain Jiang Member of Technical Staff, Verda
4:25 - 4:40 PMHiCache Room · Full & Expo Pass
Making Frontier Open Models Fast on Hopper with SGLang
Ho-Ren (Jack) Chuang Senior AI Infrastructure Engineer, ByteDance
4:10 - 4:25 PMDecode Room · Full & Expo Pass
Fast Tokens, Slow Agents: Building Open Models for Real Agent Workloads
Lucas Atkins CTO, Arcee AI
4:25 - 4:40 PMDecode Room · Full & Expo Pass
Deep Discovery for Future Prediction
Simon Shaolei Du Associate Professor, University of Washington & Lead Scientist, Apodex
4:30PM
4:30 - 4:50 PMMain Stage · Full Pass Only
Introducing GDPx by SemiAnalysis
Dylan Patel Founder & CEO, SemiAnalysis
4:40 - 4:55 PMRadixAttention Room · Full & Expo Pass
SGLang Router
Kangyan Zhou Member of Technical Staff, RadixArk
4:55 - 5:10 PMRadixAttention Room · Full & Expo Pass
SpecForge: Training the Drafts Your Server Will Actually Serve
Mao Cheng Member of Technical Staff, RadixArk
4:40 - 4:55 PMHiCache Room · Full & Expo Pass
Designing SGLang for Many Accelerators
Alex Nails Member of Technical Staff, RadixArk
4:55 - 5:10 PMHiCache Room · Full & Expo Pass
Optimizing Multimodal AI Pipelines at Scale
Boxin Jiang Tech Lead Manager, Cortex AI, Snowflake
4:40 - 4:55 PMDecode Room · Full & Expo Pass
Training Systems for Formalizing All of Mathematics
Tobias Gessler Member of Technical Staff, AxiomRishi Malhotra Member of Technical Staff, AxiomMath
4:55 - 5:10 PMDecode Room · Full & Expo Pass
ApprenticeBench: Can Agents Own an Entire Job via Learning on the Job?
Xiang Deng Co-Founder & CTO, NeoCognition
4:45 - 5:00 PMOpen Stage · Full & Expo Pass
Scaling Embodied Intelligence: Co-designing Algorithms and Infrastructure
Yuanqing Wang RLinf Core Maintainer, Peking University
5:00PM
5:10 - 5:25 PMRadixAttention Room · Full & Expo Pass
Hy-Nomas: Towards Free-Form Interactive World Generation
Ji Hou Principal Researcher, Tencent HunyuanWeizhi Wang Staff Researcher, Tencent
5:10 - 5:25 PMHiCache Room · Full & Expo Pass
Day Zero on Vera Rubin: SGLang, Miles and FlashInfer
Jingfan Sun Software Engineering Manager, CUDA Kernels & Libraries, NVIDIAYang Xu Sr. Manager, Deep Learning Software, NVIDIA
5:10 - 5:25 PMDecode Room · Full & Expo Pass
Scaling RL for Diffusion Models
Sangwu Lee Head of Research, Krea

Map

Upper level of Gateway Pavilion: sponsor booths by tier (Platinum, Gold, Silver, Star), food, beverage and lounges around the atrium, with stairs to the lower level

Lower level of Gateway Pavilion: Main Stage, Open Stage, RadixAttention, HiCache, Prefill and Decode rooms, Media Room, GPU Expo, registration, swag pick-up and box lunch

Fort Mason campus map: Gateway Pavilion and Cowell Theater on Pier 2, The Storehouse in Building D and the Museum in Building C

Pricing

Full Pass

Sold out

Developer & Education

$399

~~$849~~

$679Standard

Main Stage Talks
Breakout Rooms
Hands-On Labs
Expo Floor

Expo Pass

Developer & Education

$239

~~$499~~

$389Late Bird

Breakout Rooms
Hands-On Labs
Expo Floor

Get Expo Pass

For group passes, contact us. For sponsor employee discounts, first check with your company’s event team or colleagues; if you can’t find your discount, contact us.

Previous Events

FAQ

What’s the difference between the Full Pass and Expo Pass?

Both passes include breakout sessions, hands-on workshops, the expo floor and complimentary meals and refreshments. With an Expo Pass, you can watch the Main Stage livestream on screens at the venue. A Full Pass also gives you access to the Main Stage room. Full Passes are sold out.

On Day 1, Expo Pass entry begins when the Expo Hall opens. The morning program takes place on the Main Stage only.

What’s happening on the Main Stage and in the breakout rooms?

Main Stage talks and panels cover frontier models, AI hardware, recursive self-improvement and agents. Breakout sessions go into the technical details of inference, training and building with SGLang and Miles, with examples from production systems, multimodal AI, robotics and science.

Are discounts available?

Developers, students and researchers: Developer & Education rates are listed in Pricing.
Sponsor employees: Ask your company’s event team or colleagues for your employee discount, or contact us.
Startups: Special discounts are available to support early-stage teams.

Will food be provided?

Yes. Breakfast, lunch and refreshments are complimentary with either pass, including Kyoto Matcha, Hedge Coffee and tanghulu.

Will sessions be recorded?

Selected sessions will be recorded and shared as edited clips after the event.

Can I get a refund or transfer my ticket?

Refunds are available upon request up to three days before the event. Tickets are non-transferable.

Where is the event?

Fort Mason Center for Arts & Culture
2 Marina Blvd, San Francisco, CA

The venue is approximately 30 minutes from San Francisco International Airport (SFO) by car, depending on traffic.

Is parking available?

Paid parking is available on-site at Fort Mason Center for $5 per hour, up to $25 per day. Spaces are limited, so we recommend taking Muni or rideshare.

Still have questions? Email us.

SGLANGSUMMIT2026

Frontier Infrastructure, Silicon, Models, and Applications, on One Stage

Expo Pass

Two days of technical talks, hands-on workshops and connections with the frontier AI community.

Technical Breakouts

AI in Production

SGLang & Miles

Agentic Engineering

Training & Post-Training

AI Hardware

Real-Time AI & Media

AI for Science

Robotics

Main Stage Livestream

Hands-On Workshops

Meet the Community

Speakers

LIP-BUTAN

LILIANWENG

SOUMITHCHINTALA

ARAVINDSRINIVAS

VAMSIBOPPANA

BILLJIA

RUOMINGPANG

UJVALKAPASI

LUISCEZE

COLINBRACE

LIAMFEDUS

ERICZELIKMAN

MARKSAROUFIM

YUANDONGTIAN

LINDENLI

DYLANPATEL

NANCYZHENG

JOEYCONWAY

BEHROOZGHORBANI

YUNJIN

RICHARDGONG

KEVINFLANSBURG

SYONASARMA

NITINPERUMBETI

PARTH A.PATEL

EDWARDHU

ALEXSHAW

LIANMINZHENG

YINGSHENG

BANGHUAZHU

Sponsors

Agenda

Frontier Infrastructure, Silicon, Models, and Applications, on One Stage

Day 1 Thu, Nov 12

SGLang

SGLang Ecosystem & Miles

Fireside Chat with Vamsi Boppana

Fireside Chat with Lilian Weng

Panel: Frontier Models

SGLang for Agents on NVIDIA Vera Rubin

Fireside Chat with Soumith Chintala

SGLang for the Agentic Era: Powering Coding Agents and Automated Optimization

Fast Models, Diverse Workloads: Optimizing SGLang for Intel GPUs and Xeon CPUs

Nebius Advanced GPU Inference Optimization with Nebius: Hardware, Compression, Kernels, and Serving

LMSYS Local AI Coding Lab

Panel: Recursive Self-Improvement

The Gateway to the Agentic Enterprise

Scaling Frontier Model Training at Osmosis

Prefill Here, Decode There: Disaggregating Inference Across Datacenters and Hardware Classes

DFlash and Beyond: How Inco Runs the Fastest Open-Model Endpoints

High-Efficiency, Production-Ready Inference

Aristotle: Formal Verification, Today

Building What's Next: How Trainium and SGLang Are Shaping the Future of AI Infrastructure

Agents That Write Megakernels: Multi-Day Campaigns for GPU Performance in SGLang

Scaling SGLang on AMD Hardware with ROCm and Open Source

Training Reliable Customer Service Agents with Large-Scale RL

Tinker API Training with Miles

Training One Policy Across Many Agent Harnesses: Agentic RL Post-Training with SGLang, Miles and Harbor

The Path to Inference Efficiency

Beyond Multimodal: Building Full-Modal Intelligence with Open Weights

500 Tokens per Second Is the New Slow

Memory Accelerated Prefix Serving for HiCache Tiering

Building Labs That Learn

Technical
Breakouts

Main Stage
Livestream

Hands-On
Workshops

Meet the
Community

LIP-BU
TAN

LILIAN
WENG

SOUMITH
CHINTALA

ARAVIND
SRINIVAS

VAMSI
BOPPANA

BILL
JIA

RUOMING
PANG

UJVAL
KAPASI

LUIS
CEZE

COLIN
BRACE

LIAM
FEDUS

ERIC
ZELIKMAN

MARK
SAROUFIM

YUANDONG
TIAN

LINDEN
LI

DYLAN
PATEL

NANCY
ZHENG

JOEY
CONWAY

BEHROOZ
GHORBANI

YUN
JIN

RICHARD
GONG

KEVIN
FLANSBURG

SYONA
SARMA

NITIN
PERUMBETI

PARTH A.
PATEL

EDWARD
HU

ALEX
SHAW

LIANMIN
ZHENG

YING
SHENG

BANGHUA
ZHU