AI-Fused Prediction for Shift-Left Design Automation

The shift-left paradigm in Electronic Design Automation (EDA) seeks to move physically aware analysis into earlier design stages to improve quality, reduce iteration cost, and enable predictive digital twins. A key challenge is determining where early prediction is both accurate and impactful. This lecture presents AI-fused shift-left methodologies across critical stages of the EDA flow, from pre-RTL design to signoff, covering both digital and analog circuits. Examples include physically aware timing models integrated into logic synthesis, macro connectivity–aware optimization for improved routability, fast timing estimation tightly coupled with ECO processes for accelerated closure, and large language model–assisted analog circuit sizing. The lecture concludes with opportunities and challenges in accuracy, generalization, and toolchain integration, offering practical insights into how AI-enabled prediction can reshape modern design automation workflows.