AMD Launches 'Ross' AI Agent for Embedded Development, Claims Significant Time Reduction in HLS Optimization
AMD has released Ross, an AI agent tailored for embedded systems development, integrating with Vivado and Vitis tools to automate workflows and reportedly cut optimization time from days to hours.
ImportanceMaterialEvidenceE2 unreplicatedWrite-upQuick
US semiconductor company AMD recently launched "Ross," an agentic AI assistant designed to accelerate the entire embedded system development lifecycle, from architecture design to deployment. Named after Xilinx founder Ross Freeman, this tool signifies AMD's deep integration of generative AI into its ecosystem for FPGAs, adaptive SoCs, and x86 embedded processors.
Unlike general-purpose code generation models, AMD Ross is built upon AMD's extensive engineering knowledge base. It comprises four core components: a Model Context Protocol (MCP) server connecting to tools, an AMD-verified vector knowledge base, reusable agentic skills authored by experts, and runnable design examples. Supporting natural language interaction, it can search documentation, execute commands, and guide debugging across hardware partitioning, High-Level Synthesis (HLS), timing analysis, and low-power optimization.
In demonstration cases, AMD claims that Ross reduced the time required for a Vitis HLS optimization task on an image sensor pipeline algorithm from two days for human developers to just two hours, while also decreasing programmable logic resource usage by 35%. Developers can access the service via AMD.com/Ross. AMD plans to add more embedded tool support monthly, with power design and AI inference features expected in the fourth quarter.