In a Databricks Auto Loader job, the cloudFiles.schemaLocation option specifies a directory where Auto Loader stores the inferred schema. By persisting the schema here, Auto Loader can accurately track schema evolution over time, identify new columns, and safely recover the schema state across pipeline restarts. (Note: Checkpoint offsets are stored in a separate checkpoint directory, though often managed closely together).