David AI records and licenses channel-separated conversational audio datasets for training speech-to-speech, multilingual, and voice interaction models. It pays a network of contributors to hold natural conversations, records each speaker on a separate microphone channel, and tests each dataset design against model training outcomes before scaling collection. The company addresses a shortage of real conversational audio that captures turn-taking, interruptions, and overlapping speech, the behaviors full-duplex voice models need to learn.

David AI records and licenses channel-separated conversational audio datasets for training speech-to-speech, multilingual, and voice interaction models.

David AI

Full Memo

In-depth resources

See open roles

Similar Companies

Stay in touch