Enabling physical AI applications such as autonomous vehicles and robotics is a challenging problem due to multiple factors including data collection, model architecture design, and performing real-time inference. In this half-day tutorial we will focus on lessons and challenges encountered in developing state-of-the-art software and hardware solutions. These include tools like Isaac Sim, Isaac Lab, manipulation models such as GR00T and ACT, and hardware such as NVIDIA Jetson Thor for performing real-time inference when deploying models on real robots.
Through this tutorial, the speakers will demonstrate how attendees can create an end-to-end robotics pipeline that involves data capturing and annotation both in simulation and the real world, fine-tuning models, and eventually deploying these models on robotic systems for real-time inference. Attendees will receive deployment guides for each component of this pipeline along with slides, training datasets, and code repositories for take-home exercises.
The robotics and Physical AI space has been a strong and growing topic at CVPR, especially with computer vision advancements in VLM and VLA models that have become key research areas in recent years. The community has developed several Vision-Language-Action models such as GR00T, π0, OpenVLA, SmolVLA, and ACT.
However, building a complete robotics pipeline, from data collection to model training to deployment, remains a challenging multi-disciplinary endeavor. Data collection often requires expensive hardware and software solutions, which has prohibited many researchers from pursuing this path. Foundation models require careful architecture design and post-training strategies. And deploying models on edge devices demands hardware-aware optimizations to achieve real-time performance.
This tutorial bridges these gaps by providing a hands-on, end-to-end walkthrough of the full Physical AI stack. By the end, attendees will understand the high-level frameworks, tools, and open-source community activities around robotics, embedded devices, and model training, enabling researchers, industry partners, and communities worldwide to improve collaborations in this complex and growing field.
Johnny Núñez
Mitesh Patel
Yuqi Xie
Johnny Núñez
Mitesh Patel, Yuqi Xie, Jonathan Stephens, Dhruv Diddi
Developer Advocate
NVIDIA
Johnny is a developer advocate at NVIDIA focusing on Physical AI and Robotics. He brings experience in computer vision, edge computing, and robotics from his experience in Computer Vision and Robotics, especially on Human-Robot-Object Interactions at the University of Barcelona. He is a key member of the Jetson Research Lab driving AI and robotics on edge devices.
Sr. Developer Advocate Manager/Principal Engineer
NVIDIA
Mitesh is a Senior Developer Advocate Manager at NVIDIA. His team creates workflows for GPU-accelerated data science and Generative AI applications. He previously was a Senior Research Scientist at FXPAL and Yahoo! Labs. He holds a PhD in Robotics from the University of Technology Sydney.
Developer Advocate Manager
NVIDIA
Raymond Lo, currently based in the Silicon Valley, is the developer advocate manager at NVIDIA focusing on robotics and embedded system. Previously, he was the global lead of the Intel AI evangelist team at Intel focusing on the OpenVINO™ toolkit. Ray holds a PhD in Computer Science from the University of Toronto.
Sr. Technical Product Marketing Manager
NVIDIA
Chitoku Yato is Senior Technical Product Marketing Manager of NVIDIA Jetson Edge AI platform, where he is responsible for ensuring the best developer experience on the platform. He also works closely with the vibrant NVIDIA developer community to help evangelize the use of pre-trained AI models, developer SDKs on Jetson devices. Chitoku holds a Computer Engineering bachelor’s degree from UC Santa Barbara.
Director of Product Management for Robotics
NVIDIA
Spencer Huang is Director of Product Management at NVIDIA leading robotics software product. His work centers on open-source simulation frameworks for robot learning, synthetic data generation methodologies, and advancing robot autonomy—from industrial mobile manipulators to generalist humanoid robots. Before joining NVIDIA, Spencer earned a technical MBA from NYU with a focus on artificial intelligence.
Member of Research Staff
NVIDIA
Yuqi Xie is a Member of Research Staff at the GEAR team, where he works on the GR00T models. His research focuses on advancing robot learning through imitation learning, leveraging large-scale human and robot data.
Chief Evangelist
Lightwheel
Jonathan is Chief Evangelist at Lightwheel, working at the intersection of synthetic data, 3D content, and robotics. He partners with developers, researchers, and industry teams to bring high-fidelity simulation and physical AI workflows from prototype to real-world deployment.
Hands-on workshop: train & deploy an SO-101 arm with Isaac and GR00T
NVIDIA's reference application for robotic simulation
Robot learning framework built on Isaac Sim
Open foundation model for generalist humanoid robots
HuggingFace's open-source robotics toolkit
Ultimate platform for Physical AI at the edge
Open-source library for optimizing LLM inference
