Intel Powered AI#

Here you will find guidance that covers the frameworks, models, and tools used to build and optimize robot perception and intelligence workloads.

AI Toolkits#

AI Frameworks are the runtimes and toolkits that train, optimize, deploy, and serve models on the development kit. Each framework targets OpenVINO, so inference runs across the Intel CPU, GPU, and NPU and integrates with the ROS 2 and control stack.

Gen AI

Serve large language models on the robot for natural-language command interpretation, task planning, and reasoning, offline and on-device.

clickable cards

Vision AI

Train, optimize, and run the perception models that let a robot detect objects, segment scenes, and inspect for defects.

clickable cards

Physical AI

Develop and run the robot-learning and embodied-AI policies that map perception to action.

clickable cards

Developer Tools#

Use these tools to develop, optimize, and profile AI workloads for Robotics AI Suite applications. Install versions supported by the relevant application or Blueprint; do not combine independently pinned toolkit versions without validating the resulting environment.

Tool	Use
OpenVINO	Optimize and deploy deep-learning inference workloads.
Intel oneAPI Toolkits	Develop and profile heterogeneous C++, SYCL, and data-parallel workloads.
OpenVINO Physical AI Runtime	Accelerate your OpenVINO-powered deployment with a unified API for connecting cameras, robots, and policy inference.
Intel Physical AI Studio	Train and depoy VLA models with an easy-to-use imitation learning dataset generation platform.