Airbyte's sync process can be described in two main steps: extraction and loading. During extraction, data is pulled from the source and transformed into raw JSON format. For destinations that support normalization, SQL is executed to transform the raw JSON back into a relational table. Internally, Airbyte leverages dbt to create the SQL used for normalizing and/or transforming data. The data synchronization and normalization process is shown in this [source](https://airbyte.com/tutorials/full-data-synchronization). However, the specific details of how sync works internally depend on the source and destination connectors being used, as well as the sync mode (Full Refresh or Incremental). The sync process is managed by the Airbyte Protocol, which defines the messages and configurations used during the sync operation. You can learn more about the Airbyte Protocol in the [documentation](https://docs.airbyte.com/understanding-airbyte/airbyte-protocol). If you're looking for more information on how sync works internally, I recommend exploring the [Airbyte documentation](https://docs.airbyte.com/) and [tutorials](https://airbyte.io/tutorials) for a deeper understanding of the process.