While Voice AI is all the rage now, it wasn't a hot sector in 2017. After Dylan graduated from YC, VCs rejected him. He couldn't raise a round. They all assumed Google would do it. So he raised what he could from angels and made it work for the next 3 years.He's now built the world's most accurate Speech AI model. He's grown to 5,000 customers and raised $115M in venture capital. Last quarter, he raised a $50M Series C from Accel. Just this week,  Assembly launched Universal-1, their most powerful speech recognition model to date. Trained on over 12.5 million hours of multilingual audio data, Universal-1 is 22% more accurate than APIs from Azure/AWS/Google and has 30% fewer hallucinations than competing models.In this episode, we go through how Dylan came up with the idea, how he saw Gen AI coming long before others, and what he did in the early days to grow to $1M in ARR.