In July 2025, Meta has bolstered its artificial intelligence (AI) ambitions by recruiting top researchers from OpenAI, DeepMind, and Anthropic, alongside appointing a new leader for its Superintelligence efforts. This strategic move, led by CEO Mark Zuckerberg, aims to enhance Meta’s AI capabilities and compete with industry giants. Below is an overview of the roles these new hires, including former Scale AI CEO Alexandr Wang, are expected to play at Meta.
Alexandr Wang, previously the CEO of Scale AI, now serves as Meta’s Chief AI Officer, heading the newly formed Meta Superintelligence Labs. His role involves overseeing the development of advanced AI systems, leveraging his experience in data labeling and AI infrastructure to drive Meta’s push toward superintelligence. Trapit Bansal, a pioneer in reinforcement learning from OpenAI, will focus on enhancing reasoning models, bringing his expertise from the o-series to Meta’s AI stack. Shuchao Bi, co-creator of GPT-4o voice mode, will lead efforts in multimodal AI, building on his OpenAI background to integrate voice and visual capabilities.
Huiwen Chang, who co-created GPT-4o’s image generation at OpenAI, will work on advancing Meta’s text-to-image technologies, drawing from his Google Research innovations like MaskGIT. Ji Lin, with experience in building GPT-4o and operator reasoning, will contribute to optimizing Meta’s AI performance and reasoning frameworks. Joel Pobar, returning from Anthropic with over a decade at Meta, will enhance inference systems, applying his expertise in HHVM and machine learning to improve efficiency.
Jack Rae, a former DeepMind pre-training lead for Gemini, will spearhead training processes for Meta’s large language models, aiming to boost their reasoning capabilities. Hongyu Ren, OpenAI’s post-training expert, will refine Meta’s AI models post-development, ensuring high performance across applications. Jiahui Yu, with a background in perception and multimodal AI from OpenAI and DeepMind, will lead efforts to enhance sensory and cross-modal functionalities.
Johan Schalkwyk, a former Google Fellow with voice tech experience, will focus on integrating AI into Meta’s consumer products like WhatsApp, improving personalization. Pei Sun, another DeepMind veteran, will support pre-training efforts, leveraging his work on Waymo’s perception models. Shengjia Zhao, co-creator of ChatGPT and synthetic data expert from OpenAI, will drive data innovation, helping Meta create robust training datasets for future models.
Together, this team is set to transform Meta’s AI landscape, with Wang’s leadership guiding the charge toward superintelligence. Their diverse skills aim to address gaps in Meta’s previous models like Llama 4, positioning the company as a leader in the AI race.
Key Roles at Meta
1. Alexandr Wang: Chief AI Officer, leading Meta Superintelligence Labs with a focus on advanced AI systems.
2. Trapit Bansal: Enhancing reasoning models with reinforcement learning expertise.
3. Shuchao Bi: Leading multimodal AI, integrating voice and visual features.
4. Huiwen Chang: Advancing text-to-image technologies with image generation skills.
5. Ji Lin: Optimizing AI performance and reasoning frameworks.
6. Joel Pobar: Improving inference systems with machine learning expertise.
7. Jack Rae: Spearheading pre-training for large language models.
8. Hongyu Ren: Refining AI models through post-training processes.
9. Jiahui Yu: Leading perception and multimodal AI development.
10. Johan Schalkwyk: Integrating AI into consumer products for personalization.
11. Pei Sun: Supporting pre-training with perception model experience.
12. Shengjia Zhao: Driving synthetic data innovation for robust AI training.