Autonomous Agentic Design for Photonics (arxiv.org)

A groundbreaking approach to designing photonic devices has emerged, utilizing large language models (LLMs) to automate the process. This innovative method instructs LLMs to tackle complex photonic design problems while leveraging numerical simulations and performance criteria, enabling autonomous design loops that iterate proposals for optimal device performance. The researchers demonstrated this capability through four categories of photonic chip design, including passive components like waveguide bends and active devices such as silicon microring modulators. The significance of this development for the AI/ML community lies in its potential to drastically enhance the efficiency and effectiveness of engineering design processes, paving the way for rapid prototyping in photonics. By integrating LLMs with various design considerations—from geometric constraints to fabrication rules—the method showcases a versatile approach that can be generalized across multiple domains requiring numerically-driven evaluation. The successful combination of layout, charge transport, optical modes, and RF electrode design into a single silicon photonic modulator highlights the transformative possibilities of AI in advancing photonic technology.