264001
eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJid2V2ZW50cy1wdWJsaWMtYXBpIiwiYXVkIjpbInB1YmxpYy13ZWIiXSwic3ViIjoiaHR0cHM6Ly93d3cuZ29vZ2xlY2xvdWRldmVudHMuY29tIiwiaWF0IjoxNzc3MzgzODI1LCJuYmYiOjE3NzczODM4MjUsImV4cCI6MTc3NzM4NzQyNSwiY29udGV4dCI6eyJwYXlsb2FkIjp7ImlkIjoiIiwicmVnVHlwZU5hbWUiOiIiLCJldmVudF9pZCI6IjI2NDAwMSIsImNfMzI5MDQxNiI6IiJ9fX0.JjF5L30fCy3xBzDDYApp5b4UKvdIEyttImHwrfVvRyE
phase 7 playlists session info modal
Day 3 – April 24, 2026
Breakouts
BRK3-029 • Kubernetes, Cloud Runtimes • Advanced Technical
Platform engineering for AI: Architect a unified stack on GKE
location_on Oceanside C
schedule 8:30 AM - 9:15 AM
Siloed AI model development wastes resources and slows innovation. The future of AI platform engineering is a unified substrate. Learn to architect a multi-tenant AI training platform on Google Kubernetes Engine (GKE) that’s optimized for large-scale workloads on specialized Tensor Processing Unit (TPU) and GPU hardware. In this session, we’ll demonstrate how to use MultiKueue and a wider Kubernetes ecosystem for sophisticated job queuing and quota sharing to efficiently manage and scale AI workloads across a fleet of clusters. Dive deep into dynamic resource allocation and topology aware scheduling to maximize performance. And leave with a reference architecture for a unified, multi-cluster AI operating system. Read more
This agenda widget contains logic for handling same sessions available at multiple times
GTM Tracking Header name present
Session
contains logic for gtm tracking when modal open