Soumith Chintala, the co-creator of PyTorch and a pivotal figure in shaping modern AI infrastructure, has joined Mira Murati’s startup, Thinking Machines Lab. Chintala made the move shortly after leaving Meta, where he spent over a decade leading the development of PyTorch, the open-source deep learning framework widely adopted across industry and academia. Chintala confirmed his new role on social media, posting on X: “Thinking machines…the people are incredible.” He updated his professional profiles to reflect his position at the startup, marking his first career step since departing Meta earlier this month. Thinking Machines Lab, launched by Murati in February following her exit from OpenAI, positions itself as an AI research and product lab focused on advancing human-AI collaboration. Since its inception, the company has aggressively recruited top talent from Meta, OpenAI, Anthropic, Google DeepMind, and leading universities. Chintala’s arrival strengthens the startup’s technical leadership. His expertise in AI frameworks and scalable infrastructure is seen as crucial as Thinking Machines pushes toward developing practical tools for AI customization and deployment. The company has already attracted notable figures such as John Schulman, co-lead of ChatGPT’s development; Alec Radford, a pioneer in reinforcement learning; and Bob McGrew, former Chief Research Officer at OpenAI. Thinking Machines has been competing fiercely in the AI talent war. Federal H-1B filings reviewed by Business Insider reveal that the company has offered salaries up to $500,000 for technical roles, with multiple employees earning between $450,000 and $500,000 before bonuses and equity. The startup raised a massive $2 billion seed round earlier this year at a $10 billion valuation and is now reportedly in talks to raise additional funding at a $50 billion valuation, according to Bloomberg. Chintala’s departure from Meta comes amid sweeping changes in the company’s AI division. Meta has restructured its AI efforts into a new Superintelligence Labs unit led by former Scale AI CEO Alexandr Wang. Yann LeCun, Meta’s chief AI scientist and founder of the FAIR research group, is reportedly preparing to leave the company as well. In a reflective farewell post, Chintala praised PyTorch’s global impact, noting it is now used in production at nearly every major AI company and taught in institutions from MIT to rural India. He said he felt the project had matured enough to continue without his day-to-day leadership and that he was ready to embrace something new and challenging. While Thinking Machines has not yet released its full product lineup, its first tool, Tinker, enables users to fine-tune large language models more easily. It is already being used by researchers at Princeton and Stanford, as well as early business clients. The company has faced some leadership turnover, including the departure of cofounder Andrew Tulloch, who returned to Meta in October, according to The Wall Street Journal. Thinking Machines has not commented on Chintala’s hiring, and he did not respond to requests for further details.