Applications ranging from search engines to virtual assistants need access to accurate, timely knowledge about entities and their relationships. To enable the development of such applications, Apple researchers developed a knowledge graph (KG) construction and serving platform that continuously integrates billions of facts about real-world entities from structured databases and unstructured datasets. Saga ingests and transforms data from both batch and streaming sources into ontology-aligned source data, which is used to update or extend a stable knowledge graph and live knowledge graph. The Graph Engine, based on a federated polystore, stores the stable KG, computes knowledge views over the graph and exposes query APIs for use cases like text search and analytics. The live graph query engine is optimized for low-latency graph search queries over a live KG constructed from sources that contribute temporal facts in real-time. Machine learning enables fact ranking and verification, missing fact imputation, and named entity recognition and disambiguation.