Design Automation: Active-Passive Co-Design

Building on our work in passive-network inverse design, we are developing surrogate modeling, inverse design, and agentic design flows that jointly optimize active and passive components for high-speed analog and RF ICs.

Human-AI collaboration appears to be the optimal strategy for this effort. Guided by circuit intuition and domain expertise, the designer narrows the vast search space to a promising, physically realizable subspace of topologies and dimensions. AI then explores this feasible subspace, jointly optimizing active and passive components through iterations that a human designer would find tedious. Our goal is not to ask AI to generate an analog/RF IC entirely from scratch, but to translate expert circuit reasoning into an executable optimization problem. This “divide-and-conquer” approach preserves interpretability while allowing designers to focus on architectural innovation rather than routine parameter tuning.