Conception and Software Implementation of a Nuclear Data Evaluation Pipeline

We discuss the design and software implementation of a nuclear data evaluation pipeline applied for fully reproducible neutron-induced cross sections $^{56}$Fe above resolved resonance region using model code TALYS combined with relevant experimental data. The emphasis is on mathematical technical aspects not $^{56}$Fe, which tentative. building blocks employed in are discussed detail. A unified representation data, systematic statistical errors, parameters defects enables application Generalized Least Squares (GLS) its natural extension, Levenberg-Marquardt (LM) algorithm, large collection LM algorithm tailored to accounts exact non-linear physics determine best estimates quantities. Associated uncertainty information derived from Taylor expansion at maximum posterior distribution. also terms IT (=information technology) blocks, such as those efficiently manage retrieve EXFOR library distribute computations scientific cluster. Relying we elaborate sequence steps perform evaluation, retrieval correction uncertainties marginal likelihood optimization (MLO) after screening thousand -- including Gaussian process priors energy dependent fitting about 150 algorithm. manual Dockerfile simplified installation available www.nucleardata.com.