What's new for AI on GKE: Training, serving, and agents
Google kubernetes engine Gke Ai infrastructure Machine learning Model serving Reinforcement learning Ai agents Google cloud next Mlops Kubernetes
This Google Cloud Next 26 session covers GKE's new AI capabilities released over the past year, focusing on training, serving, reinforcement learning, and agentic workloads. The video demonstrates how GKE has evolved to support complex AI infrastructure needs with improved reliability, scalability, and cost-effectiveness. It includes real-world case studies from Thinking Machines Lab and Baseten. Ideal for platform engineers, MLOps practitioners, and organizations running AI workloads on Google Cloud.
Related videos
Orchestrating ML/AI workloads with TPUs on GKE
Shipping custom models at scale from fine-tuning to inference | BRK234
Run AI at scale with Ray + Kubernetes using Anyscale on Azure | ODSP914
Evolving KServe: The Unified Model Inference Platform for Both Predictive and... F. Spolti & J. Lee
Beyond VLLM: Distributed LLM Inferencing With Llm-d on Kubernetes - Ravindra Patil, Red Hat
Implementing DeepMind innovation: Deep research API
Engineering the future of Kubernetes for AI at scale
Under the hood for startups: How Google DeepMind makes modeling decisions
Post-Training and Deploying Open Source Reasoning Models in Foundry | DEM321
Deploy. Observe. Learn. Reinforcement learning for production agents | BRK231
CNCF On-Demand: Autonomous Agents on K8s – Durable Execution for AI
