Siemens EDA uses NVIDIA NeMo Gym to improve its chip design agent

Siemens EDA used NVIDIA NeMo Gym to check the engineering files its chip design agent produced, trace errors, and refine the agent's skills and tool descriptions, according to NVIDIA AI. On Siemens EDA's standard cell characterization benchmark, NVIDIA says the changes raised the agent's composite verification score from about 55% to 95%, with 66% fewer agent turns and 64% fewer output tokens.

NVIDIA published a techblog describing what the team measured and changed, and how other engineering teams can apply the approach. The figures come from NVIDIA's account and have not been independently verified.