Databricks explains how its AI Research team uses AI Runtime for GPU training

Databricks says its AI Research team uses AI Runtime (AIR) to train models ranging from a few billion to nearly a trillion parameters on NVIDIA H100 and B300 GPUs.

AIR runs preflight checks, monitors node health, and cordons off failing nodes so training jobs avoid problematic infrastructure. The team also uses the AIR CLI, custom Docker containers, MLflow tracking, and Unity Catalog FUSE mounts to iterate quickly.