<p>The radioactive materials assistance has driven many contributions in diverse fields, including industrial, nuclear power plants, and medical treatments, while in fluid mechanics, these particles play a significant role in enhancing the velocity and temperature profiles. The objective of the study under consideration is to generate approximate solutions for the mathematical models representing radioactive materials contaminated in MHD blood flow involving slip and nonlinear convection using deep layered recurrent neural networks supported by Bayesian distributed optimization (LRNN-BDO). The development of the existing model is constructed by transforming PDEs into a system of nonlinear ODEs through suitable variations. Analysis is performed with multi-class parameters in the system model via magnetic interaction, Prandtl number, Eckert number, the linear/non-linear mixed convection, radiation, temperature, and non-dimensional nanofluid. The generated synthetic data is numerically formulated by exploiting an implicit backward differentiation scheme in the presence of uranium dioxide (UO₂) and thorium dioxide (ThO₂), with base fluid as blood. The proposed solutions of the LRNN-BDO consistently align with the reference solutions that the errors are approximately equals to zero, exhibiting robust, efficient, and reliable networks performance evaluated through various assessment metrics corresponds to iterative convergence on mean squared error, adaptive controlling measures of optimization, error frequency distribution on histograms and regression statistics for exhaustive validation for radioactive materials contaminated MHD blood flow model involving slip and nonlinear convection.</p>