Skip to main content

Building AI-Enhanced Robotics for Real-World Environments

abstract background

AI-powered robots are moving from controlled demos to real-world deployment. But what does it take to make them reliable, scalable, and able to handle physical AI workflows? Join industry leaders and innovators as they discuss how AMD solutions can solve workflows in factories, warehouses, and beyond. This session explores the importance of balanced system architecture, open collaboration, and trends shaping physical AI, from models and simulation to agentic workflows and multimodal reasoning

July 23, 2026 13:00 - 13:45

Amey Deosthali
Amey Deosthali

Head of Robotics, Industrial, and Healthcare, Embedded Computing Group | AMD

Topic

Physical AI & Robotics

Session Type

Breakout Session

Related Sessions

View All Sessions
  • Tool Calling Robots: Agentic Workflows Complementing Physical AI

    Robots with tool calling capabilities expose coarse- and fine-grained control as software APIs. A robot with access to these APIs can then reason over what it sees, adjusting cameras, checking for humans, or sequencing actions – all before any motion executes. Join us for an interactive discussion to have all your questions answered on how this agentic layer, running on capable edge hardware, can greatly amplify the emerging benefits of physical AI.;Robots with tool calling capabilities expose coarse- and fine-grained control as software APIs. A robot with access to these APIs can then reason over what it sees, adjusting cameras, checking for humans, or sequencing actions – all before any motion executes. Join us for an interactive discussion to have all your questions answered on how this agentic layer, running on capable edge hardware, can greatly amplify the emerging benefits of physical AI.

    July 23, 2026

  • Introduction to Robotics with ROS2 on AMD

    This intermediate, interactive session focuses on deploying an end-to-end Vision-Language-Action (VLA) pipeline using ROS 2 and multimodal models to control a robotic arm. Participants will work through the steps to build and deploy a system that responds to physical inputs with real-world actions. Prior familiarity with ROS 2 and VLA concepts is recommended.;This intermediate, interactive session focuses on deploying an end-to-end Vision-Language-Action (VLA) pipeline using ROS 2 and multimodal models to control a robotic arm. Participants will work through the steps to build and deploy a system that responds to physical inputs with real-world actions. Prior familiarity with ROS 2 and VLA concepts is recommended.

    July 23, 2026

  • Always-On Physical AI Applications with AMD Ryzen AI Software

    This beginner session covers deploying vision-detection models to the Ryzen™ AI NPU using ONNX Runtime. Attendees will examine how offloading inference to the NPU optimizes system resource allocation, freeing CPU and iGPU capacity for other workloads in a physical AI pipeline.;This beginner session covers deploying vision-detection models to the Ryzen™ AI NPU using ONNX Runtime. Attendees will examine how offloading inference to the NPU optimizes system resource allocation, freeing CPU and iGPU capacity for other workloads in a physical AI pipeline.

    July 22, 2026

  • Build Physical AI Kernels for Vision Applications with AMD ROCm

    This beginner to intermediate session introduces kernel-level development using ROCm. Participants will build a kernel application in C++, analyze performance using AMD profiling tools, and apply optimization techniques. Prior familiarity with kernel programming is recommended.;This beginner to intermediate session introduces kernel-level development using ROCm. Participants will build a kernel application in C++, analyze performance using AMD profiling tools, and apply optimization techniques. Prior familiarity with kernel programming is recommended.

    July 22, 2026

  • Enabling Physical AI on AMD

    In this beginner-to-intermediate session, users will execute multimodal vision and language models for robotics on the Ryzen™ AI iGPU via ROCm and Python.The session covers model loading and examines how vision and language pipelines integrate to produce actionable robot behavior.;In this beginner-to-intermediate session, users will execute multimodal vision and language models for robotics on the Ryzen™ AI iGPU via ROCm and Python.The session covers model loading and examines how vision and language pipelines integrate to produce actionable robot behavior.

    July 22, 2026

  • Real2Sim for Embodied Data Generation

    This beginner-to-intermediate workshop introduces a practical pipeline for building scalable embodied Synthetic Data generation systems for robotics and physical AI. Participants will learn how to transform real-world objects into interaction-ready simulation assets through a Real2Sim workflow focused on object-centric reconstruction, and simulation integration on AMD Instinct GPUs.;This beginner-to-intermediate workshop introduces a practical pipeline for building scalable embodied Synthetic Data generation systems for robotics and physical AI. Participants will learn how to transform real-world objects into interaction-ready simulation assets through a Real2Sim workflow focused on object-centric reconstruction, and simulation integration on AMD Instinct GPUs.

    July 22, 2026

  • Build Physical AI Kernels for Vision Applications with AMD ROCm

    This beginner to intermediate session introduces kernel-level development using ROCm. Participants will build a kernel application in C++, analyze performance using AMD profiling tools, and apply optimization techniques. Prior familiarity with kernel programming is recommended.;This beginner to intermediate session introduces kernel-level development using ROCm. Participants will build a kernel application in C++, analyze performance using AMD profiling tools, and apply optimization techniques. Prior familiarity with kernel programming is recommended.

    July 22, 2026

  • Enabling Physical AI on AMD

    In this beginner-to-intermediate session, users will execute multimodal vision and language models for robotics on the Ryzen™ AI iGPU via ROCm and Python.The session covers model loading and examines how vision and language pipelines integrate to produce actionable robot behavior.;In this beginner-to-intermediate session, users will execute multimodal vision and language models for robotics on the Ryzen™ AI iGPU via ROCm and Python.The session covers model loading and examines how vision and language pipelines integrate to produce actionable robot behavior.

    July 22, 2026