Using AI-driven wavefront-shaping for scalable optical circuit switching (Conference Presentation)

The demand for efficient, scalable, and high-capacity interconnects has surged in the rapidly evolving landscape of data center networks. While the current optical interconnect technologies offer benefits over electrical interconnects, they fall short in the scalability of the port count in a functionalized way, a crucial aspect for AI applications. We introduce a novel reconfigurable optical interconnect device, which stands distinct as it harnesses the principles of free space diffraction coupled with programmable spatial light modulation techniques optimized with an AI-driven algorithm. This approach enables various switch modalities, including unicast, multicast, and wavelength switching, making it adaptable to future network architectures and implementations without requiring any changes in the physical layer. This innovative combination paves the way for designing more compact, energy-efficient, and low-latency network architectures, which are imperative for meeting the computational demands of today's data centers.