architecture

A system design practice note on separating real-time LLM serving from asynchronous feedback collection at 20K peak TPS

A system design note on asynchronous document processing, GPU workers, queues, and storage choices for Document AI