The deep scattering layer (DSL) is a ubiquitous feature across the global ocean composed of a diverse group of organisms, collectively known as mesopelagic micronekton, who play key roles in carbon cycling and serve as important pelagic prey resources. Yet, the fine-scale vertical structure and dynamics of the DSL remain largely understudied. The DSL in the northern Gulf of Mexico was examined using shipboard and autonomous wideband echosounders. An unsupervised machine learning framework was applied to wideband backscatter data to identify unique morphological groups (echo-types) and assess their distribution during the daytime within the DSL. Echo-type dynamics were further explored during diel vertical migration using an image processing algorithm to segment migrating layers and assess heterogeneity within and across them. Multivariate and network analyses provided strong support for depth and site-specific patterns, potentially linked to predation risk, reproduction, life-history traits, physicochemical conditions, or oceanographic forcing. Next, to complement this fine-scale view, multifrequency shipboard echosounders and depth-stratified net catch data were integrated in a Bayesian probabilistic echo solving model to produce a comprehensive deep-pelagic time series. Together, these approaches provide new insights into the vertical complexity of mesopelagic communities and new avenues to examine the deep-pelagic at the micro and macroscale level. Moreover, this work underscores the value of integrating multiple data sources with shipboard echosounders and utilizing novel modeling approaches, to further disentangle the ecological interactions, life history strategies, and oceanographic features contributing to mesopelagic community structure.