In this research, we investigate the segregation of Gaussian light beam wavelengths using corrugated triangular wave-index interfaces with the beam propagation method (BPM). We present a novel approach for optical wavelength demultiplexing (DEMUX) utilizing the Goos-Haenchen (GH) shift and continuous refraction at the interface. By exploiting wavelength-dependent lateral shifts (GH shifts), reflected beams achieve spatial beam separation. Light beams exhibit distinctive "turning points" in the medium, resulting in spatially separated "back-refracted" beams. Meta-heuristics were employed to optimize operational parameters for this demultiplexing optical splitter, enabling its application in "corrugated triangular wave-index integrated-optics" structures within optical fiber networks. Artificial Intelligence (AI) in conjunction with an Ant Colony (AC) was utilized for optimal beam incidence. Findings suggest improved demultiplexer performance with three media compared to two media for 2 and multi-wavelength demultiplexing, highlighting a significant academic contribution of this study.