PyTorch Conference: October 22–23, 2025, San Francisco, CA
Open Source AI Week’s flagship event, PyTorch Conference 2025, brings together AI pioneers, researchers, developers, and startup founders for keynotes, technical talks, tutorials, and community programming focused on open source AI. Expect deep dives from maintainers and ecosystem leaders, plus practical sessions on kernels, compilation, inference, distributed training, and real-world AI applications.
Building on last year’s success, this year’s program expands with dedicated co-located events on Tuesday, October 21, and the launch of PyTorch training and certification. Use this guide to navigate the week, plan your schedule, and get the most from your time onsite.
Quick links:
At a Glance
⭐ Tuesday, October 21
- Open Agent Summit
- Measuring Intelligence Summit
- AI Infra Summit
- PyTorch on the Edge
- Training
- Startup Showcase
⭐ Wednesday, October 22
- Keynotes & Breakouts, BoFs, Meet the Experts
- Expo Hall & Gear Store
- Llama Palooza
- Flare Party & Poster Presentations
⭐ Thursday, October 23
- Keynotes & Breakouts, BoFs, Meet the Experts
- Expo Hall & Gear Store
- Women & Non-Binary in PyTorch Lunch
Tuesday, October 21: Co-located Events and Training
Start your week with focused summits and training opportunities. See the event website for the latest information on registration and fees.
Co-located events
- Open Agent Summit: 8:30 AM, Room 2005–2007
- Measuring Intelligence Summit: 8:30 AM, Room 2001–2003
- AI Infra Summit: 1:00 PM, Room 2001–2003
Other co-located programming
- Startup Showcase: AI startups pitch live to top VCs. Your chance to network with industry leaders, discover breakthrough ideas, and get inspired! Starts at 5:00 PM, Room 2014–2016.
- PyTorch Associate Training: Instructor-led course covering core competencies for the PyTorch Certified Associate (PTCA) exam. Starts at 8:30 AM, Room 2004. Participants receive a PTCA exam voucher upon completion.
Evening social events
- Arm at PyTorch: Connect, Chat, Chill – 6:30 PM, One Yerba Buena
Open Source AI Week
PyTorch Conference 2025 Schedule Highlights
Wednesday, October 22
- Morning keynotes (Room 3000–3012)
- Welcome & Opening Remarks – Matt White (PyTorch Foundation)
- Recognizing the PyTorch Foundation CICD Infrastructure Community – Eli Uriegas (Meta)
- vLLM & DeepSpeed Updates – Simon Mo (vLLM / UC Berkeley) and Tunji Ruwase (Snowflake)
- Robert Nishihara & Ion Stoica (Anyscale / UC Berkeley)
- The Physical Turing Test: Solving General Purpose Robotics – Jim Fan (NVIDIA)
- Sponsored Keynote: From Transactions to Transformers – Hardware Built for Enterprise AI – Christian Jacobi (IBM)
Nathan Lambert (AI2)
Session highlights
- AI Engineering – Chip Huyen (Tep Studio)
- Efficient MoE Pre-training on AMD GPUs with TorchTitan – Liz Li & Yanyuan Qin (AMD); Matthias Reso (Meta)
- ExecuTorch 1.0: GA for Mobile and Embedded Applications – Mergen Nachin (Meta)
Afternoon keynotes (Room 3000–3012)
- Animashree (Anima) Anandkumar (Caltech)
- Sponsored Keynote: Build AI Anywhere with ROCm Software – Anush Elangovan (AMD)
- PyTorch Technical Deep Dive
- Sponsored Keynote: Unlocking AI at Scale for PyTorch Developers – Jeff Gehlhaar (Qualcomm)
- Panel Discussion: Agentic AI – Aparna Dhinakaran (Arize AI); Jerry Liu (LlamaIndex); Adam Jones (Anthropic); Samuel Colvin (Pydantic)
Community
- Poster Sessions & Flare Party – Sponsored by neptune.ai
Thursday, October 23
Morning keynotes (Room 3000–3012)
- Welcome Back – Matt White (PyTorch Foundation)
- Reclaiming Simplicity: From PyTorch’s Thousand Ops to Groq’s Tensor Ops – Andrew Ling (Groq)
- Percy Liang (Stanford / Together)
- Signal vs. Noise: Betting on AI’s Next Wave – Sarah Catanzaro (Amplify Partners); Sarah Guo (Conviction); Brian Zhan (CRV); Peter Deng (Felicis); moderated by Joe Spisak (Meta)
- Robot Foundation Models – Sergey Levine (UC Berkeley)
Session highlights
- Transformers: Standardizing Model Definitions Across the PyTorch Ecosystem – Lysandre Debut & Arthur Zucker (Hugging Face)
- Helion: A High-Level DSL for Kernel Authoring – Jason Ansel (Meta)
- Monarch: A Distributed Execution Engine for PyTorch – Michael Suo (Meta)
Afternoon keynotes (Room 3000–3012)
- Dawn Song (UC Berkeley)
- Jeremy Howard (fast.ai) in conversation with Anna Tong (Forbes)
- Panel Discussion: Hardwares & Accelerators – Dylan Patel (SemiAnalysis); Sharon Zhou (AMD); Peter Salanki (CoreWeave); Nitin Perumbeti (Crusoe Energy); moderated by Mark Saroufim (Meta)
- Eric Xing (MBZUAI / CMU / Genbio AI)
- Community Awards & Closing Remarks – Matt White (PyTorch Foundation)
Track Overview
Following the morning keynotes, breakout sessions at PyTorch Conference 2025 are organized into thematic tracks that reflect the breadth of innovation across the PyTorch community. Each track explores a distinct area of AI research, development, or application, helping you focus on the topics most relevant to your work.
Tracks include:
Deep Learning Compilers, Kernel Authoring & Accelerators
Go deep into PyTorch extensibility with compilers, custom kernels, and hardware acceleration.
Who should attend: Developers, systems engineers, hardware specialists.
What you’ll learn: Extending PyTorch, optimizing kernels, and maximizing accelerator performance.
Edge
Discover how PyTorch powers AI on mobile, embedded, and IoT devices.
Who should attend: Engineers building edge, mobile, and IoT AI solutions.
What you’ll learn: Deployment strategies, model compression, and on-device optimization.
Foundation Projects / Ecosystem
Get the latest on vLLM and DeepSpeed, plus ecosystem projects from Hugging Face, Lightning, and community contributors.
Who should attend: PyTorch users, contributors, and ecosystem developers.
What you’ll learn: Updates, roadmaps, and integration best practices for PyTorch ecosystem tools.
LLMs
Dive into the latest breakthroughs in large language models with PyTorch.
Who should attend: Researchers, ML engineers, developers, product teams.
What you’ll learn: Architectures, training optimizations, and real-world LLM applications.
PyTorch at Scale
Learn how to build, train, and serve models at massive scale.
Who should attend: ML practitioners and infrastructure engineers.
What you’ll learn: Distributed training, parallelism, and scaling AI in production.
PyTorch in Academia
See how PyTorch is advancing research and education across disciplines.
Who should attend: Researchers, students, educators.
What you’ll learn: Research highlights, teaching practices, and academic collaboration with PyTorch.
PyTorch in Industry
Explore real-world use cases of PyTorch driving business innovation.
Who should attend: Industry practitioners, product leaders, technical decision-makers.
What you’ll learn: Deployment strategies, lessons from production, and success stories from the field.
Reinforcement Learning
Discover the latest in RL algorithms, frameworks, and real-world applications.
Who should attend: Researchers and engineers exploring RL.
What you’ll learn: RL techniques in PyTorch for robotics, gaming, and optimization.
Data
Explore the future of intelligent data pipelines that shape, curate, and enhance AI training.
Who should attend: Data scientists, ML engineers, researchers.
What you’ll learn: Tools and methods for smarter data pipelines and adaptive datasets.
Onsite Activities
Make the most of your time at PyTorch Conference 2025 with these community and networking opportunities throughout the week. Explore the full list
Keynote Session Raffles
Wednesday, October 22 – Thursday, October 23, Room 3000–3012
Attend daily Keynote Sessions for a chance to win prizes. Pick up a raffle ticket when you enter and stay through the session. Winners are announced at the end. Must be present to win.
Birds of a Feather (BoFs)
Wednesday, October 22 & Thursday, October 23, Exhibit Hall, BoF Lounge
Join informal, topic-driven discussions with your peers on everything from model deployment to ethical AI. No sign-up required.
- Wed 10:30–11:00 AM: Strong Model, Weak Product? Let’s Fix That
- Wed 3:20–3:50 PM: Deploying LLMs at the Edge: Lessons from the Field
- Thu 10:30–11:00 AM: Applying DevSecOps Lessons to MLSecOps
- Thu 3:20–3:50 PM: Ethical and Responsible AI: PyTorch’s Framework for Safer Model Deployment
Meet the Experts
Wednesday, October 22 & Thursday, October 23
Engage directly with PyTorch engineers and maintainers in small group discussions. No registration needed.
- Wed 10:30–11:00 AM: Developers of PyTorch Compiler
- Wed 3:20–3:50 PM: Developers of PyTorch Dev Infra
- Thu 10:30–11:00 AM: PyTorch Module Maintainers
- Thu 3:20–3:50 PM: vLLM Maintainers
Book Signings
Expo Hall
Meet leading authors, grab a signed copy, and learn from their work.
- Wed 3:20–3:50 PM: Chip Huyen
- Thu 3:20–3:50 PM: Sebastian Raschka
PyTorch Gear Store
Wednesday, October 22 – Thursday, October 23, Exhibit Hall
Pick up exclusive PyTorch hoodies, hats, bags, and more. Open during Exhibit Hall hours.
Llama Palooza!
Wednesday, October 22, 3:20–3:50 PM
Meet the llamas — pre-trained on PyTorch, fine-tuned for selfies. Don’t forget to tag your photos with #PyTorchCon.
Poster Presentations
Wednesday, October 22, 6:00–7:30 PM, Exhibit Hall
Connect with presenters and explore cutting-edge research in this interactive evening session.
Women & Non-Binary in PyTorch Lunch
Thursday, October 23, 12:35–1:25 PM, Room 2022–2024
A networking lunch for attendees who identify as women or non-binary. Sponsored by Google.
Social Media Contest
Share your PyTorch Conference experience for a chance to win exclusive PyTorch gear.
How to enter
Post on LinkedIn or X during the conference or any of this week’s co-located events.
Include the hashtags #PyTorchCon and #PyTorchLive.
Tag @PyTorch (@PyTorch on X and @PyTorch on LinkedIn).
What we’re looking for
Insightful takeaways, creativity, and photos or video that capture your experience.
Prizes & announcement
Five winners will receive PyTorch gear, announced during the closing keynote at 3:00 PM PT in Room 3000–3012. Winners must be present to receive prizes.
Open Source AI Week: Oct 18–26, 2025
PyTorch Conference anchors Open Source AI Week, a week-long celebration of open collaboration in AI across the Bay Area. The schedule includes meetups, hackathons, summits, and socials exploring the intersection of AI, ML, and open source.
A few event highlights:
Oct 18–19: Synthetic Data AI Agents Challenge
Oct 20: AMD AI DevDay, Real Time Video AI Summit, Meta-Qualcomm-GitHub Edge Event, Intuit Open Source Meetup
Oct 21: Measuring Intelligence Summit, Open Agent Summit, Triton Developer Conference, Elastic{ON} AI, AI Infra Summit, Startup Showcase
Oct 22–23: PyTorch Conference
Oct 24: Women in Data Science (IBM Silicon Valley Lab), dAGI Summit, GPU MODE IRL Hackathon
See You in San Francisco
PyTorch Conference 2025 is where the next generation of open AI comes to life. Explore the PyTorch Conference website and Open Source AI Week website to plan your week. We can’t wait to welcome you onsite!
In the meantime, check out PyTorch Conference 2024 Highlights