Dynamo KVBM - Managing Memory at Scale

Description

Got questions about KV cache management in your LLM deployment? Join us for a session on Dynamo's KV Block Manager (KVBM)- a system that decouples memory management from run-time logic to enable memory-efficient, high-throughput LLM systems. We’ll dive into architecture, integration with LLM inference runtimes, integration with NIXL, and our roadmap. Come ready to chat, drop your questions in the comments, and learn more from the engineers building KVBM. Be part of the journey and contribute on github: https://bit.ly/dynamoKVBM