A data-driven strategy for virtual material analysis and synthesis enables the representation, characterization, and generation of solid electrolyte interphase (SEI) configurations based on kinetic Monte Carlo (KMC) simulations. A variational autoencoder (VAE) model, equipped with a property predictor, learns key features of 2D SEI configurations from selected samples. The model analyzes essential features at the bottleneck to assess how properties like thickness, porosity, density, and volume fraction influence learned data-driven characteristics. To improve classification, inputs to the VAE are conditioned with a reaction barrier set linked to specific SEI conditions, allowing for the generation of SEI configurations with customized physical properties.