Ciliary movement holds substantial importance in the realm of biomedical research. Disorders arising from impaired ciliary motion, such as primary ciliary dyskinesia and specific respiratory ailments, underscore the critical need to investigate both the normal functioning and dysfunction of cilia. Such research is fundamental to the development of effective treatments and therapies. This work explores the application of Levenberg–Marquardt backpropagation neural networks (LMBP-NNs) for simulating the flow of a non-Newtonian Jeffery fluid in peristaltic motion. Inspired by synchronous ciliary movement, microfluidic systems mimicking cilia functionality have been developed. These artificial cilia can generate controlled fluid flow patterns, enabling various applications such as drug delivery, mixing, and particle sorting in lab-on-a-chip devices. The present study focuses on measuring thermal and mass transfer in beating cilia during multiphase peristaltic pumping of Jeffery fluid in two- and three-dimensional channels using transverse magnetohydrodynamics in a porous medium. The governing equations for the non-Newtonian Jeffery fluid in both the fluid and dusty phases are formulated, and exact solutions are calculated for fluid phase velocity, particulate phase velocity, temperature, concentration, and pressure gradient. The physical behavior is investigated by graphing dimensionless parameters for velocity, temperature, and concentration profiles in both solid and liquid phases. The trapping phenomena is depicted by showing streamlines of various dimensionless characteristics such as porosity velocity and volume percentage. The LMBP-NNs approach is then used to train, test, and verify the neural network models using the reference datasets. The accuracy of the LMBP-NN is assessed using statistical data such as mean square error, curve fitting graphs, regression plots, and error histograms. The study also investigates flow model limits for momentum, temperature, and concentration profiles using visual representations. This study illustrates the computational capability of LMBP-NNs in modeling the peristaltic flow of Jeffery fluid and gives useful insights into fluid flow behavior in microfluidic devices containing artificial cilia. The findings help to better understand and optimize these systems for a variety of fluid manipulation and transportation applications.