Building Multi-Cloud GenAI Platforms Without the Pain
🎥 Recorded live at the MLOps World | GenAI Summit 2025 — Austin, TX (October 9, 2025) Session Title: Building Multi-Cloud GenAI Platforms Without the Pain Speaker: Romil Bhardwaj, Co-creator, SkyPilot Talk Track: LLMs on Kubernetes Abstract: As GenAI workloads continue to expand, organizations are hitting hard limits with single-cloud strategies — from GPU shortages to escalating costs and vendor lock-in. In this talk, Romil Bhardwaj, Co-creator of SkyPilot, presents a hands-on guide to building multi-cloud AI platforms that unify compute resources across hyperscalers, neoclouds, and on-prem clusters — without adding operational chaos. Romil demonstrates how to seamlessly connect and orchestrate workloads across AWS, GCP, Azure, Coreweave, Nebius, and Lambda through a single abstraction layer. You’ll see practical scheduling strategies for GPU allocation, automated cloud selection for cost optimization, and techniques for handling cross-cloud data movement and dependency management. This session provides a roadmap for ML teams to secure GPU capacity, reduce costs, and focus on building great AI products instead of fighting infrastructure complexity. What you’ll learn: • Why going multi-cloud is becoming essential for GenAI workloads • How to unify compute across clouds and clusters with SkyPilot • Strategies for workload scheduling and cost optimization • How to manage cross-cloud data movement and dependencies • Practical patterns for scalable, resilient multi-cloud AI platforms.
