AI use cases

Prefix caching

Reuse KV-cache memory for requests with shared system prompts and avoid repeated prefill computation.

From How KV Cache Speeds Up LLMs for Faster AI Models on GPUs by IBM Technology