The increasing demand for compact and energy-efficient photonic systems operating in the green spectral region has accelerated the development of integrated wavelength-division multiplexing (WDM) solutions. However, conventional multiplexing approaches rely on discrete and bulky optical components, leading to increased insertion loss, limited scalability, and reduced integration capability. At the same time, emerging neuromorphic photonic architectures require multifunctional platforms capable of simultaneous wavelength multiplexing and optical weighting. In this work, we propose an AI-assisted design and optimization framework for a four-channel green-light WDM multiplexer based on multicore polymer optical fiber (AI-POF technology). The device employs a multicore structure with embedded polycarbonate (PC) cores arranged in a symmetric hexagonal lattice within a fluoropolymer matrix. Passive wavelength multiplexing is achieved through engineered inter-core coupling, enabling efficient operation across the 500–560 nm spectral range without the need for external gratings or lenses. Artificial intelligence– enhanced RSoft beam propagation method (BPM) simulations are utilized to optimize the coupling geometry and switching length, allowing a single coupling region to satisfy phase-matching conditions for all four wavelength channels simultaneously. The optimized 20-mm fiber segment demonstrates low insertion loss (0.13–0.55 dB), uniform 20-nm channel spacing, and strong resilience to thermal and wavelength variations. The AI-driven optimization further enables precise control of wavelength-dependent power distribution, supporting passive optical weighting analogous to synaptic functionality in neuromorphic systems. Beyond multiplexing, the proposed architecture allows adaptive tuning of the switching length to achieve controlled weighted summation at the output, effectively integrating WDM functionality with photonic computing capabilities. Experimental validation using a two-channel prototype at 500 nm and 540 nm confirms strong agreement with simulation results, demonstrating efficient spatial multiplexing and signal convergence within the fiber. The presented AI-POF-based design eliminates the need for multiple coupling sections and discrete optical elements, significantly reducing structural complexity while enhancing spectral efficiency. This work establishes a scalable and compact platform for green-band WDM systems and AI-driven neuromorphic photonic interconnects, highlighting the potential of combining polymer photonics with intelligent design methodologies.

Dror Malka received his BSc and MSc degrees in electrical engineering from the Holon Institute of Technology (HIT) in 2008 and 2010, respectively, Israel. He has also completed a BSc degree in Applied Mathematics at HIT in 2008 and received his Ph.D. degree in electrical engineering from Bar-Ilan University (BIU) in 2015, Israel. Currently, he is a Senior Lecturer in the Faculty of Engineering at HiT. His major fields of research are nanophotonics, super-resolution, AI silicon photonics and fiber optics. He has published around 80 refereed journal papers, and 80 conference proceedings paper.