Recorded at the IT Press Tour in Boston, this episode explores the fascinating story behind Paradigm4 and how a challenge in large-scale biomedical research ultimately led to the creation of flexFS, a cloud-native filesystem designed to tackle some of today's biggest data infrastructure challenges.
Joining me on the podcast is David Freund from Paradigm4, who shares how the company was originally founded to help scientists work with enormous datasets in fields such as genomics, bioinformatics, and precision medicine. As researchers began working with population-scale datasets such as the UK Biobank, the team discovered that existing storage technologies either couldn't deliver the performance they needed, lacked the functionality required, or became prohibitively expensive at scale. Our conversation explores the moment Paradigm4 realized it would need to build its own solution, why traditional approaches to cloud storage often struggle under modern analytics workloads, and how flexFS emerged from a real-world customer problem rather than a technology trend. David also explains why object storage has become such an attractive foundation for modern infrastructure, while discussing the challenges of latency, performance, and cost that still need to be addressed.
We also discuss why many organizations investing heavily in AI infrastructure may be overlooking one of the biggest constraints on performance. While much of the industry conversation focuses on GPUs and compute power, David argues that data access, movement, and management are becoming equally important considerations as AI workloads continue to grow.
Along the way, we touch on cloud independence, resilience, large-scale analytics, and why flexibility across cloud providers is becoming an increasingly important requirement for enterprise technology leaders. Whether you're working in AI, life sciences, cloud infrastructure, or enterprise data management, this episode offers an interesting perspective on how customer problems can sometimes lead to entirely new categories of technology.
Could the next major AI bottleneck be data rather than compute? And are organizations paying enough attention to the infrastructure feeding their most important workloads? I'd love to hear your thoughts.