Vsevolod Vikulin, Yandex (head of the quick answers service)
This is a simple, honest talk, similar to an LLM system design interview. The author shares his three years of experience deploying LLMs and breaks it down into five "stops" of an imaginary LLM bus: why do we need an LLM, where to start, how to evaluate quality, how to choose an architecture, how to save money.
Continue reading — it's free
Google or a one-time email code. No password.
