Edge AI

Running inference near the data source to cut latency and uplink traffic, using NPUs, GPUs or accelerator modules on an embedded host.

  • Multi-protocolEdge AI, Arm architecture, Embedded Linux, Yocto Project
  • Multi-manufacturer35 manufacturers in this selection
  • Built-in documentationDatasheets and linked resources
  • SupportHelp choosing and integrating

Architecture

Typical Edge AI architecture

From the field to the uplink: each role of the chain and the product families that fill it.

Request a quote
  1. 2. Accelerator

    Add-in module for inference.

    See the products

This is an architecture pattern, not a certified reference design. Verify every interface, rating and certification in the datasheets.

Products

Products for a Edge AI solution

Request a quote

Industries

Sectors that use these product families

All industries

Technologies and protocols

Compatibilities and standards

All technologies

Documentation and resources

Guides and documentation

  • Manual xpi-s905x3Manual · Geniatech Open
  • Manual xpi-3576Manual · Geniatech Open
  • Manual xpi-3568Manual · Geniatech Open
  • Manual xpi-3566-zeroManual · Geniatech Open
All resources

Before choosing

What to verify

Model format and toolchain supported by the accelerator
TOPS figures compared at the same precision
Power and thermal envelope
Update path for models in the field