AI-Assisted EM Structure Synthesis Using a Software-Defined Open-Source EDA Flow

The recent success of AI/ML across many fields highlights the importance of workflows that are natively script-driven. In analog design, and especially in RF/mmWave, traditional EDA tools remain closed-source and GUI-based, making them difficult to customize and integrate with AI-assisted methods for design, optimization, and analysis. The experience-driven nature of RF/mmWave design further compounds this challenge, but with the right software interface, AI can play a transformative role. This work introduces a software-defined, open-source EDA flow that unifies parametric layout generation, electromagnetic (EM) simulation, and circuit-level analysis within a single Python environment. By abstracting layout generation and EM–circuit co-simulation into fully programmable interfaces, the flow enables large-scale data generation and integration with AI/ML wrappers. Examples are presented showing how different AI techniques can be wrapped around the flow to automatically synthesize and optimize complex EM structures. Both theoretical considerations, such as which classes of models align best with electromagnetic design tasks, and practical aspects of embedding AI into a live EDA workflow are discussed. The presentation concludes with a real-time demonstration of AI-assisted EM structure synthesis, illustrating how software-defined design flows can lower barriers to AI adoption in RF/mmWave design and unlock new opportunities for increased productivity through automation.