

From GPUs-as-a-Service to Workloads-as-a-Service: Flex AI’s Path to High-Utilization AI Infra
18 snips Sep 28, 2025
Brijesh Tripathi, CEO of Flex AI, combines his rich background in AI and HPC architecture to revolutionize AI infrastructure. He discusses the burdens of DevOps that slow down small AI teams and highlights Flex AI's innovative workload-as-a-service approach. Brijesh breaks down the challenges of accessing heterogeneous compute, the importance of consistent Kubernetes layers, and how to smooth costs for spiky workloads. He also shares insights on handling real-time vs. best-effort workloads, maximizing utilization, and ensuring that AI teams can focus on creativity instead of complexity.
AI Snips
Chapters
Transcript
Episode notes
Supercomputer Handover Sparked Flex AI
- Brijesh Tripathi described building the Aurora supercomputer and the long handover to scientists.
- That experience inspired Flex AI to simplify access to compute for researchers and developers.
DevOps Friction Slows Small Teams
- Small teams waste time on infrastructure instead of product because DevOps complexity is high.
- Flex AI aims to remove that burden so teams can iterate on models faster.
Stabilize Kubernetes, Allow BYO Containers
- Standardize the Kubernetes layer so developers don't handle cloud-specific library drift.
- Offer bring-your-own-containers to capture edge cases while keeping a stable orchestration base.