Generates synthetic German personas grounded in population and geographic distributions for model training and synthetic data generation. The dataset contains 1 million records with six German-language persona types, based on census and other official statistics.