A high-throughput and memory-efficient inference and serving engine for LLMs
AICI: Prompts as (Wasm) Programs