In this work, a coarse-grained model is developed for highly cross-linked bisphenol A diglycidyl ether epoxy resin with diaminobutane hardener. In this model, all conformationally relevant coarse-grained degrees of freedom are accounted for by sampling over the free-energy surfaces of the atomic structures using quantum mechanical simulations. The interaction potentials between nonbonded coarse-grained particles are optimized to accurately predict the experimentally measured density and glass-transition temperature of the system. In addition, a new curing algorithm is also developed to model the creation of highly cross-linked epoxy networks. In this algorithm, to create a highly cross-linked network, the reactants are redistributed from regions with an excessive number of reactive molecules to regions with a lower number of reactants to increase the chances of cross-linking. This new algorithm also dynamically controls the rate of cross-linking at each local region to ensure uniformity of the resulting network. The curing simulation conducted using this algorithm is able to develop polymeric networks having a higher average degree of cross-linking, which is more uniform throughout the simulation cell as compared to that in the networks cured using other curing algorithms. The predicted gel point from the current curing algorithm is in the acceptable theoretical and experimental range of measured values. Also, the resulting cross-linked microstructure shows a volume shrinkage of 5%, which is close to the experimentally measured volume shrinkage of the cured epoxy. Finally, the thermal expansion coefficients of materials in the glassy and rubbery states show good agreement with the experimental values.