System-Level Simulation: Towards Hardware–Software Co-Design for Photonics-Enabled AI Accelerators

Realizing the promised gains of photonic AI requires moving beyond component-level metrics toward end-to-end evaluation. By linking differentiable circuit models with automatic architecture synthesis and system-level co-simulation, we quantify the critical interplay between circuit parameters, accelerator parameters, and memory configurations. This holistic approach will enable us to demonstrate a photonic computing advantage, providing a definitive roadmap for designing energy-efficient, high-throughput accelerators capable of meeting the demands of next-generation AI workloads.