Building the Voice OS at Willow
Building a Voice OS requires moving beyond accurate ASR to mastering latency, personalization, and style inference. The talk details using LLMs and RL to structure raw speech into polished, usable communication.
Read article