RESEARCH · RESEARCH · #1976
Siemens Fuse EDA agent reliability raised from ~55% to ~95% using NVIDIA NeMo Gym
Siemens used NVIDIA NeMo Gym to evaluate and improve its Fuse EDA AI Agent for standard-cell characterization, boosting a composite verification score from about 55% to ~95% across a representative benchmark. Refining agent skills and Model Context Protocol (MCP) tool descriptions based on verifier diagnostics also cut agent turns from 155 to 53 and reduced output tokens from ~81K to ~29K.
KEY POINTS
- Siemens used NVIDIA NeMo Gym to evaluate and improve its Fuse EDA AI Agent for standard-cell characterization, boosting a composite verification score from about 55% to ~95% across a representative benchmark.
- Refining agent skills and Model Context Protocol (MCP) tool descriptions based on verifier diagnostics also cut agent turns from 155 to 53 and reduced output tokens from ~81K to ~29K.
- This demonstrates a reproducible agentic evaluation workflow that diagnoses failure modes across multistage chip-design pipelines, materially improving reliability and efficiency needed for production adoption.
WHY IT MATTERS
This demonstrates a reproducible agentic evaluation workflow that diagnoses failure modes across multistage chip-design pipelines, materially improving reliability and efficiency needed for production adoption.