IntelligentHQ
IntelligentHQ
Skip to content

Edge AI & Embedded Computing

Computers and software for AI outside the data center: Jetson modules for robots and devices, IGX for industrial and medical systems with functional safety, Holoscan for real-time sensor processing, and JetPack.

Technology profiles for this category are in research.

Overview

Some AI has to run next to the sensor. A robot cannot wait on a network round trip, a medical device may not send data off site, and a remote site may have no link at all. This category covers NVIDIA hardware and SDKs for that setting.

Jetson is a family of compact modules, from Orin Nano to Jetson Thor, programmed with JetPack 7. IGX is an industrial-grade platform for robotics, medical devices and instruments; IGX Thor includes a Functional Safety Island designed to meet ISO 26262 and IEC 61508, and IGX includes AI Enterprise with up to 10 years of support. Holoscan SDK is an open-source runtime for real-time processing of streaming sensor data on IGX and other aarch64 or x86 systems. DeepStream adds video pipelines on Jetson.

The audience is embedded engineers, device makers and teams deploying AI across many sites.

If latency, privacy and connectivity are not constraints, running the model in the cloud is usually easier to update and monitor. Very small tasks, such as wake-word detection, may not need a GPU module.1234

Problems it addresses

  • Latency in control loops1

    Machines that act on what they sense need local inference. Jetson targets robots, drones and vision systems.

  • Safety certification2

    Industrial and medical systems need functional safety. IGX Thor includes a safety island designed to meet ISO 26262 and IEC 61508.

  • Long product lifecycles2

    Devices stay in the field for years. IGX Orin 700 is listed with support through 2033.

  • High-bandwidth sensors3

    Holoscan processes streaming sensor data, such as surgical video, in real time.

A typical workflow

  1. Fix the constraints

    Set power, size, latency and certification targets.

  2. Choose the module2

    Pick Jetson for compact devices or IGX for industrial and medical systems.

  3. Build the pipeline3

    Use Holoscan for sensor streams or DeepStream for video.

  4. Fit the model

    Reduce model size and precision until it fits the module's memory and power budget.

  5. Plan fleet updates

    Design how models will be updated and monitored across all deployed devices.

NVIDIA Solution Architect

Describe your project and get an explainable architecture.

Open the lab

Next steps

  1. Write down the power budget, latency target and any certification standard before choosing hardware.

  2. Prototype on a Jetson or IGX developer kit with your real sensors.

  3. Plan how models will be updated and monitored once many devices are in the field.

  4. Use the NVIDIA Solution Architect to outline the edge and cloud split for your project.

Sources

  1. NVIDIA Jetson embedded systems (product page) (opens in a new tab)NVIDIA · Vendor-reported
  2. NVIDIA IGX (opens in a new tab)NVIDIA · Vendor-reported
  3. NVIDIA Holoscan SDK (developer page) (opens in a new tab)NVIDIA · Vendor-reported
  4. NVIDIA DeepStream SDK (developer page) (opens in a new tab)NVIDIA · Vendor-reported

Fill out the form below to request your copy.

Name(Required)