Peter Tong, also known as Shengbang Tong, joined AMI as a member of the technical staff in September 2026. He defended his PhD in computer science at NYU Courant in April 2026, with supervision from Yann LeCun and Saining Xie. He previously graduated from UC Berkeley with a triple major in computer science, applied mathematics, and statistics.

His work spans visual representation learning and multimodal systems. Selected projects include Cambrian-1, a vision-centric exploration of multimodal large language models, and Web-SSL, on scaling language-free visual representation learning.

Before his PhD, Tong conducted research at Berkeley AI Research and NYU’s Center for Data Science. He received the OpenAI Superalignment Fellowship for 2024–25, and his open-source work includes EMP-SSL, a self-supervised learning approach designed to reduce training time.