How David AI is Building the Data Layer for Audio AI

David AI, the world’s first audio data research lab, led by Tomer Cohen and Ben Wiley, has raised $50 million in Series B from Meritech, NVIDIA, and existing investors Alt Capital, First Round Capital, Amplify Partners, and Y Combinator.

Audio AI has the potential to bring artificial intelligence into everyday life, making it accessible to billions. But to reach that potential, next-generation audio interfaces need vastly more data and evaluations than exist today. David AI is building that foundation, creating the high-quality data that powers the models behind future audio interfaces, enabling real-world applications like humanoid robots, wearables, personal assistants, and generative media.

Several of the world’s leading AI labs and Mag7 companies are already advancing their audio AI models with David AI’s support. Yet even as model performance accelerates, progress is constrained by a shortage of rich, high-quality training data and robust evaluation frameworks.

Speech presents unique challenges: unlike code or text, it’s inherently subjective and context-dependent. A great customer service interaction differs entirely from a natural chat with an AI companion. Beyond words, countless variables, emotion, tone, pace, accent, environment, shape meaning. Multilinguality adds further complexity, as audio cannot be cleanly translated and dialectal nuances often defy standardization.

At the core, these are data challenges, solvable only with datasets diverse and detailed enough to capture, and generalize across, the full spectrum of human communication.