AI agents for photonic integrated circuit design automation
A. Surjalal Sharma, Yuqi Fu, Vahid Ansari, Rishabh Iyer, Fiona Kuang, Kashish Mistry, Raisa Islam Aishy, Sara Ahmad, J. Matres, Dirk Englund, Joyce K. S. Poon · APL Machine Learning · 2025
We present photonics intelligent design and optimization, a proof-of-concept multi-agent framework that converts natural-language photonic integrated circuit (PIC) design requests into layout mask files. This work demonstrates end-to-end PIC design automation using large language models (LLMs), with the goal of achieving structurally valid rather than performance-qualified layouts. We compare seven reasoning LLMs using a testbench of 102 design descriptions that ranged from single devices to 112-component PICs. The success rate for single-device designs was up to 91%. For design queries with ≤15 components, o1, Gemini-2.5-pro, and Claude Opus 4 achieved the highest end-to-end pass@5 success rates of ∼57%, with Gemini-2.5-pro requiring the fewest output tokens and the lowest cost. Future work will extend this framework toward performance qualification through expanded datasets, tighter simulation and optimization loops, and fabrication feedback integration.