Embedded engineers bounce between hardware debug, software bring-up and machine learning more than most developers do. Advanced Micro Devices thinks an AI helper can shoulder part of that load.
The company has introduced Ross, an assistant wired into its embedded toolchain. It covers field-programmable gate arrays, adaptive systems-on-chip, x86 embedded processors and edge AI platforms.
Its remit runs from splitting work between hardware and software to high-level synthesis, machine learning tuning, power analysis and schematic review.
When a build breaks, Ross is meant to explain why. It can surface timing problems, translate obscure tool messages and flag parts of a design worth revisiting.
The assistant is neutral on the model side. Engineers pick their own large language model, editor and command line while keeping ties to AMD’s software. It is built from Model Context Protocol servers, a validated knowledge base and reusable agent skills kept as open Markdown files.
Salil Raje, who leads AMD’s embedded group, said the work keeps getting harder as teams span hardware design, deployment, inference and system-level planning. One assistant, he argued, can pull those threads together.
Ross is available now, with AMD promising fresh tools and workflows each month.
