PyTorch 2025: Unlocking the Future of AI with Qualcomm Technologies
Join us at PyTorch 2025, the premier conference for developers, researchers, and innovators shaping the future of AI. This year’s event builds on the momentum of last year’s global gathering, which welcomed over 2,000 attendees from 36 countries. With an expected turnout of 3,000 participants, PyTorch 2025 promises to be a vibrant hub of collaboration, learning, and cutting-edge innovation.
Meet Team Qualcomm Technologies at Pytorch 2025
At Qualcomm Technologies, we’re reimagining the AI development journey — from cloud to edge — with a unified, developer-first approach. This year at PyTorch 2025, we’re showcasing how our AI software stack empowers developers to build, optimize, and deploy PyTorch models across a wide spectrum of devices and use cases.
Our presence at the event is anchored in the belief that AI should be scalable, efficient, and accessible. Whether you're fine-tuning large language models in the cloud or deploying multimodal applications on-device, Qualcomm’s AI platforms — including Cloud AI 100 Ultra, AI Hub, ExecuTorch, and NexaSDK — provide the tools and infrastructure to make it happen.
We’re bringing to life our vision of “Torch Once, Run Everywhere,” enabling developers to seamlessly transition from model development to real-world deployment. With a focus on hybrid AI inferencing, performance optimization, and developer productivity, Qualcomm is helping the PyTorch community unlock new levels of innovation.
Lorem ipsum dolor sit amet, consectetur adipiscing elit,
Sed ut perspiciatis unde omnis iste natus error sit voluptatem accusantium doloremque laudantium, totam rem aperiam, eaque ipsa quae ab illo inventore veritatis et quasi architecto beatae vitae dicta sunt explicabo. Nemo enim ipsam voluptatem quia voluptas sit aspernatur aut odit aut fugit, sed quia consequuntur magni dolores eos qui ratione voluptatem.
Name Lastname
Sed ut perspiciatis
Heading
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
0:35
Title
Video description
3:46
Title
Video description
PRE-EVENT HIGHLIGHTS
PyTorch Developer Hackathon (Meta & Qualcomm)
Date: Monday, October 20 at 9:00 AM-9:00PM PDT
Location: GitHub HQ 275 Brannan St, San Francisco, CA 94107
A 10-hour in-person hackathon where developers will build high-quality on-device AI applications using PyTorch + ExecuTorch on Snapdragon mobile devices. This hands-on experience will showcase the power of Qualcomm’s platforms and foster collaboration across the developer ecosystem
PRE-EVENT HIGHLIGHTS
AI Hub + AWS External Workshop
Date: Tuesday, October 21 at 1200 PM-7:00PM PDT
Location: Location: InterContinental 888 Howard Street San Francisco, CA 94103
Learn how to bring your PyTorch models to AI Hub, optimize them for deployment, and run them on real devices. Gain insights into performance metrics, model compilation, and meet the AI Hub team to explore how to elevate your AI development.
SESSION OVERVIEW
Unlocking AI at Scale for PyTorch Developers
Date: Wednesday, October 22 at 2:46pm
Location: Room 3000 - 3002
As AI continues to scale across industries and devices, developers need tools that are not only powerful but also efficient, accessible, and optimized for real-world deployment. In this session, we’ll explore how Qualcomm Technologies is enabling developer-friendly AI inference — optimized for performance, energy efficiency, and cost-effectiveness — to unlock AI at scale.
We’ll dive into the latest trends driving inference across the cloud, edge, and hybrid environments, and how these shifts are enabling new levels of personalization, privacy, and responsiveness. Learn how Qualcomm’s flexible AI software stack empowers PyTorch developers to seamlessly deploy models across billions of devices — from data centers to smartphones and beyond.
Plus, don’t miss an exciting announcement that builds on our ongoing efforts to make the Qualcomm Hexagon NPU even more accessible — unlocking its full potential for developers everywhere.
Speaker: Jeff Gehlhaar, SVP, Engineering
DEMO 1 OVERVIEW
Cloud AI Developer playground: Finetune, Compile, Inference.
Fine-tune Llama 3.1 8B model and deploy for inference on Qualcomm Cloud AI 100 Ultra using the Qualcomm AI Inference Suite. Easy to use UI/APIs for dataset & model lifecycle management through deployment
DEMO 2 OVERVIEW
Qualcomm AI Hub: Optimize and validate AI models for on-device deployment
Qualcomm AI Hub simplifies AI developer workflows by offering a comprehensive platform for model deployment, optimization, and validation on edge devices. With 100+ pre-optimized models, automatic conversion from PyTorch and ONNX, and real-time performance benchmarking, developers can bring their models to life on real Snapdragon-powered devices.
DEMO 3 OVERVIEW
Develop Real AI Apps: PyTorch and ExecuTorch with NexaSDK
With NexaSDK, developers can run PyTorch models across CPU, GPU, and NPU with a single line of code. This hands-on demo features a multimodal AI assistant and a powerful profiling toolkit that provides deep insights into inference time, bottlenecks, and performance — all on-device.
DEMO 4 OVERVIEW
Torch Once, Run Everywhere: Scalable AI
Deploy Llama and Whisper models using ExecuTorch, optimized with the QNN delegate. This demo walks through model preparation, conversion, device integration, and real-time testing — showcasing how to scale AI from development to deployment with ease.
Snapdragon and Qualcomm branded products are products of Qualcomm Technologies, Inc. and/or its subsidiaries.
Connect with our communities
Forums
Connect with fellow developers and get answers from Qualcomm Technologies’ experts. Dive in to collaborate, innovate, and find solutions—start exploring now.
YouTube channel
Find tutorials, overviews, and use cases to support your development with Qualcomm® SoCs, HDKs, tools, and SDKs
GitHub community
Open-source software that leverages the unique capabilities of today's mobile devices
Developer Discord
Get support from experts, connect with like-minded developers, and access exclusive virtual events.
