From competition student to infrastructure founder
Wang grew up in Los Alamos, where his parents worked in physics research. He competed in mathematics, physics, and computer science before entering MIT to study machine learning. What changed his path was not another competition result, but internships and engineering work that showed him many machine-learning teams lacked not algorithms, but a reliable way to obtain high-quality data continuously.
In 2016 he founded Scale AI with Lucy Guo through YC and became CEO. The earliest customers came from autonomous driving: cameras and lidar generated enormous volumes of data, and engineers needed roads, vehicles, and pedestrians labeled accurately in order to train models.
Turning “manual labor” into a system
Data labeling looks labor-intensive on the surface, but the moat is not just labor cost. How tasks are decomposed, who performs them, how they are reviewed, how errors are tracked, and how results are connected back to a customer’s training pipeline determine delivery quality and speed.
Scale turned these steps into software and, during the generative-AI cycle, moved the business toward expert data, red-team testing, and model evaluation. The form of the data changed; the underlying judgment did not: model capability depends on continuous, reliable human feedback.
The identity shift in 2025
In June 2025, Meta invested $14.3 billion in Scale AI, a transaction corresponding to a company valuation above $29 billion. Scale was not fully acquired by Meta; Wang stepped down as CEO, kept his board seat, and became Meta’s Chief AI Officer.
This was not an ordinary exit. It moved his position of power from CEO of an infrastructure company into a platform with models, compute, and distribution to billions of users.