No US grads on Meta's Superintelligence team—all hires are immigrants
What's the story
Meta, the parent company of Facebook, has hired 11 leading artificial intelligence (AI) researchers for its new Superintelligence Lab. In a notable move that highlights the importance of immigrant talent in the tech industry, all these hires are immigrants. The team includes experts from OpenAI, Google DeepMind, Anthropic, and Google Research. They have worked on advanced AI models like GPT-4o, Gemini, and Chinchilla.
Global talent
Immigrant talent in the tech industry
The new hires come from diverse academic backgrounds, with roots in India, South Africa, China, the UK, and Australia. Their work covers a range of fields including large language models (LLMs), multimodal systems, reinforcement learning, and infrastructure at scale. The team's expertise will play a crucial role in Meta's ambition to lead in artificial general intelligence (AGI).
Individual contributions
Alumni from IIT, Berkeley, Tsinghua, and Princeton
The team features Trapit Bansal, an IIT Kanpur alumnus with a PhD from UMass Amherst, known for his work in reinforcement learning and co-creating OpenAI's o-series models. Shuchao Bi, educated at Zhejiang University and UC Berkeley, contributed to GPT-4o's voice mode and o4-mini. Huiwen Chang, a Princeton and Tsinghua graduate, designed the MaskIT and Muse architectures that power generative image models.
Continued expertise
More researchers from Google DeepMind and OpenAI
Ji Lin, a graduate of MIT and Tsinghua, focuses on scaling large models, enhancing image generation, and improving GPT-4o's efficiency and deployment. Joel Pobar, educated in Australia, with experience at Anthropic, has developed core infrastructure for HHVM, Hack, and PyTorch (key tools behind large-scale AI systems). Jack Rae, originally from the UK and holding degrees from CMU and UCL, formerly led large model pretraining at DeepMind, where he played a key role in developing Gopher, Chinchilla, and Gemini 2.5.
Last few
Minds behind AI's evolution
Hongyu Ren contributed to the development of GPT-4o and the o-series mini models, focusing on enhancing model safety and robustness. Johan Schalkwyk, a University of Pretoria graduate and a pioneer in Google's speech technology, led the Maya team. Pei Sun, an alum of CMU and Tsinghua, played a key role in advancing perception systems at Waymo and in the Gemini project. Jiahuai Yu, educated at USTC and UIUC, worked on GPT-4 and multimodal systems at OpenAI and Gemini.
Information
Meta also has a co-creator of ChatGPT on board
Shengjia Zhao, a Tsinghua graduate with a PhD from Stanford, co-created ChatGPT, GPT-4, and o4-mini, with a focus on data synthesis and safety in generative AI. His contributions, along with those of his peers, have been instrumental in enhancing OpenAI's multimodal systems.