Skypilot can integrate well with ClearML, WandB, or other ML experiments tracking tools. It also supports full reproducibility with tools like DVC. Checkpointing is supported using private object storage buckets. Skypilot enables distributed training to train models in a shorter time. It also allows users to save costs by using spot instances. For re-triggering old experiments, users can easily rerun a job using the YAML file. Skypilot currently does not have a GUI.