Cross-platform high-performance engine for machine learning inference and training acceleration, supporting models from multiple frameworks with graph optimization across diverse CPU and hardware targets.