

Data Volume, Quality, and Model Degradation for AI at Scale - with Sunitha Rao of Hitachi Vantara
6 snips Oct 9, 2025
Sunitha Rao, SVP and General Manager for Hybrid Cloud Business at Hitachi Vantara, shares her expertise in modernizing data infrastructure for AI. She discusses the challenges of explosive data growth and emphasizes the importance of data quality over sheer volume. Sunitha explains how to prevent model degradation through effective monitoring and service-level objectives, while also advocating for energy-efficient data management. She provides actionable insights for aligning IT investments with sustainability goals, helping enterprises scale their AI initiatives responsibly.
AI Snips
Chapters
Transcript
Episode notes
Data Growth Breaks Traditional Infrastructure
- Explosive AI data growth creates multiple infrastructure gaps beyond just storing more bits.
- Leaders must unify data, compute, storage, networking, and orchestration to scale responsibly.
Plan Beyond Pure Compute Scaling
- Don't assume adding compute alone solves AI scaling; plan for GPU shortages and data-center power/cooling.
- Include networking and ultra-low latency links for distributed training demands.
AI Storage Demands Mixed, Elastic Design
- Storage patterns for AI differ: mixed file, block, and object needs with linear scaling and tiering.
- Legacy storage focused on one model will bottleneck read-write and lifecycle needs for AI.