# Checkpointing for Resiliency and Performance in AI Pipelines

How much does storage latency affect checkpointing in AI training? NeuralMesh delivered 40-50% lower checkpointing latency than high-performance NFS across FastPitch, ASR, and BioMegatron on AWS.

**Type:** White Paper

[Download PDF](/api/resource-pdf?slug=checkpointing-for-resiliency-and-performance-in-ai-pipelines)