David AI is the first dedicated audio data research company — it builds the training datasets that power speech and conversational AI models. Founded in 2024 by former Scale AI executives Tomer Cohen (CEO) and Ben Wiley (CTO), the company designs, records, cleans, and annotates high-quality conversational audio at research grade. Its flagship products include Converse (natural two-speaker, channel-separated English conversations) and Atlas (15+ languages with accent and dialect metadata). Customers license these datasets to train speech recognition, voice synthesis, translation, and full-duplex dialogue systems. The business model is B2B licensing — perpetual or time-limited access — sold to Fortune 100 companies, leading AI labs, and the majority of the Magnificent Seven tech companies. David AI's core thesis: as audio AI races toward its "ChatGPT moment," the real bottleneck isn't compute — it's the scarcity of rigorously curated, channel-separated speech data. The company applies the same methodological rigor to data collection that top labs apply to model development.