Fluid is an open-source cloud-native data orchestration and acceleration system that provides elastic data abstraction for AI applications running on Kubernetes. It leverages distributed caching (based on Alluxio) to reduce data access latency and improve training efficiency for machine learning workloads. The system acts as a Kubernetes operator that manages data placement and caching across cluster nodes to optimize I/O performance for AI frameworks.