# Serving Ray AIR `Checkpoint`s

[Ray AI Runtime (AIR)](air) is a scalable and unified toolkit for ML applications. Ray Serve is part of Ray AIR.
It natively integrates with AIR's `Predictor` and `Checkpoint` abstractions to enable seamless transition from training to serving.
You can learn more about how to serve Ray AIR Checkpoints in [this guide](air-serving-guide).