David AI

David AI is an audio data research company that builds the data layer for audio AI. Founded in 2024 by Tomer Cohen and Ben Wiley, both former Scale AI engineers, the San Francisco-based firm develops high-quality audio training datasets for frontier AI research, serving many of the world's leading AI labs and major technology companies.

The company's work spans audio AI, speech recognition, machine translation, speech synthesis, conversational AI, and the curation, development and evaluation of training data. Its published datasets include Converse, natural two-speaker English conversations; Atlas, spanning 15+ languages with dialect metadata; Chorus, multi-speaker conversations; and Dialog, expert domain conversations.

David AI applies an R&D-driven approach to dataset development, with a proprietary iterative process: hypothesize, design, experiment, evaluate and iterate, productionize, and release. The company has raised $80M in total - a $5M seed, a $25M Series A, and a $50M Series B led by Meritech with participation from NVIDIA - at a valuation of $500 million.