From Models to Reality: What It Takes to Run AI Reliably at Scale
Session Overview
As AI moves from experimentation into production, organizations are hitting new challenges in how data-center infrastructure is designed and operated. At scale, success is shaped not only by models and compute, but by the ability to manage growing volumes of data reliably and efficiently across real-world environments.
In production, decisions about how data is stored, accessed, and managed increasingly drive system performance, operational cost, and scalability β across both training and inference. Gaps or inefficiencies in the data layer introduce friction that limits the effective use of compute and complicates AI operations over time.
This session explores how production AI is reshaping infrastructure considerations, drawing on patterns observed across a range of cloud and data-center environments. It examines why data architecture and lifecycle management are becoming more central to AI system design β and what "AI-ready" can practically mean as AI becomes part of everyday operations rather than isolated projects.
What you'll take away
- Beyond Models and Compute β why reliable, efficient data management at scale is what truly separates AI that works in production from AI that stalls.
- The Data Layer as a Differentiator β how storage, access, and lifecycle decisions shape performance, cost, and scalability across training and inference.
- What "AI-Ready" Really Means β practical infrastructure patterns observed across global cloud and data-center operators as AI becomes part of everyday operations.
Speaker
Colin Presly, Seagate's VP of Customer Engineering, has over 25 years of hard-drive technology and customer experience. His career includes leadership roles in research and development, precision equipment engineering, the CTO office, and global customer-serving technical organizations. His work has focused on helping organizations design and operate infrastructure capable of supporting increasingly demanding workloads, including data-intensive and AI-driven systems. Through regular engagement with global cloud and data-center operators, Colin brings a perspective on how infrastructure requirements are evolving as AI moves from experimentation into production. He holds a Master of Engineering from Imperial College London.