Generate high-fidelity, privacy-preserving synthetic tabular datasets using generative models (CTGAN, TVAE, Gaussian Copula). Preserve complex marginal probability distributions, multi-column correlations, and conditional dependencies while passing strict empirical privacy audits (Wasserstein distance, mutual information similarity, and nearest-neighbor distance to original training data). Trigger when generating test datasets, sharing data with external vendors, or balancing imbalanced classes.
