vLLM on AWS - Testing to Production and Everything in Between

Explore practical architectural patterns for deploying and scaling large language models in production environments through this comprehensive conference talk from AWS re:Invent 2025. Walk through the complete journey from initial testing to production deployment, covering essential steps including model evaluation with vLLM, performance benchmarking, and optimization techniques. Learn to implement efficient autoscaling solutions using Ray and vLLM, compare different inference servers like Triton and vLLM while understanding their respective trade-offs. Dive deep into productionalization strategies using AIBrix and gain actionable insights for building robust, scalable LLM infrastructures. Designed for ML engineers and architects working on LLM deployments, this 51-minute session provides hands-on guidance for managing the complexities of large language model operations in cloud environments.