Predictive inference framework that leverages synthetic data to improve sample efficiency in data-scarce settings, while providing rigorous, distribution-free finite-sample coverage guarantees.