Cloud-Native Model Serving - vLLM's Lifecycle in Kubernetes

Explore the deployment and management of Large Language Models in Kubernetes environments through this 36-minute conference talk from DevConf.US 2025. Learn how vLLM, a leading open-source project for LLM inference serving, maximizes throughput while minimizing resource usage through its unique features including dynamic batching and distributed serving. Discover the complete lifecycle of deploying AI/LLM workloads on Kubernetes, covering seamless containerization techniques, efficient scaling strategies using Kubernetes-native tools, and robust monitoring practices to ensure reliable operations. Understand how vLLM simplifies complex AI workloads and optimizes performance to make advanced inference accessible for diverse and demanding use cases. Gain insights into integrating vLLM with Kubernetes to build reliable, cost-effective, and high-performance AI systems that drive innovation in scalable LLM deployment.