The formation of high-entropy alloy (HEA) solid solutions of simple structures (fcc, bcc, and hcp) can now be predicted with considerable accuracy using various data-driven methods. However, predicting high-entropy intermetallics (HEI) and their composite alloys that are usually associated with functional properties, e.g., half-metallic, multiferroic, magnetocaloric, thermoelectric, and topologically-nontrivial, etc., have met with challenges. Modeling HEI is often constrained by the availability of datasets and the effectiveness of machine learning features. We have developed physics-based and compound-specific models to explore the complex compositional landscape of HEIs through feature engineering and experimentation. The high prediction capability of the models enables the design of ordered BCC (B2) and Heusler (L21) HEI for high-performance properties.