Soumith Chintala is an Indian-American artificial intelligence researcher and engineer, widely recognized as a co-creator of the PyTorch open-source deep learning framework ^{[1] [2]}. During his eleven-year tenure as a Research Scientist at Meta AI, formerly Facebook AI Research (FAIR), he led the PyTorch project from its inception into a foundational tool for the global AI community ^[3]. Chintala is a prominent advocate for open-source AI development, and on November 6, 2025, he announced his plan to depart from Meta later that month to pursue new ventures ^[3].

​

​

Education

Chintala completed his undergraduate studies at the Vellore Institute of Technology (VIT) in India. He later moved to the United States for his graduate education at New York University (NYU) ^[1]. At NYU, he earned a Master of Science in Computer Science between 2010 and 2012, followed by his doctoral work in the same field from 2012 to 2016 ^[4]. His research at NYU focused on deep learning and computer vision, where he worked under the mentorship of AI pioneer Yann LeCun ^[1].

Career

Early Career

During his time at New York University, Chintala became deeply involved in the nascent deep learning community. He was a key maintainer and core developer of Torch7, a scientific computing framework based on the Lua scripting language that gained traction in prominent research labs ^[1]. Alongside Pierre Sermanet and Yann LeCun, he also co-maintained EBLearn, a C++ based deep learning library. Chintala created convnet-benchmarks, a benchmarking suite for deep learning models that was widely considered a standard by hardware manufacturers like NVIDIA, AMD, and Intel between 2015 and 2017. His early work also included influential open-source projects such as an implementation of the DCGAN architecture in Torch ^{[5] [1]}.

Meta AI (Facebook AI Research)

Chintala joined Facebook AI Research (FAIR) in May 2014 and remained with the company, later rebranded as Meta AI, for eleven years ^{[4] [3]}. He described the early, formative years of FAIR as "absolutely magical," characterized by a small, collaborative team building state-of-the-art AI in an open environment. Before his work on PyTorch, he contributed to a variety of influential projects at FAIR, including Generative Adversarial Networks (GANs), StarCraft AI bots, the first FAIR computing cluster, and object detection systems ^[3].

His most significant contribution at Meta was co-creating and leading the PyTorch framework, which was publicly released in 2016. He served in numerous capacities for the project, including project manager, technical lead, communications lead, and release engineer, guiding its growth from an internal research tool to a globally adopted platform ^[3].

Departure from Meta

On November 6, 2025, Chintala announced his intention to leave Meta, with an effective departure date of November 17, 2025 ^{[3] [2]}. He stated that his primary work on PyTorch at the company felt "now complete" and that he was ready to explore new challenges. His stated reasons for departing included a desire to "do something small again" and to avoid the "counterfactual regret of never trying something outside Meta." He also expressed confidence that the PyTorch project and its new leadership team were resilient and no longer required his direct oversight, noting, "The project didn't need me anymore" ^[3].

Industry observers viewed the announcement as a pivotal moment for the open-source AI ecosystem, interpreting it as a sign of PyTorch's maturity and the success of its community-driven governance model rather than a disruption ^[2].

PyTorch

PyTorch is Chintala's most recognized contribution to the field of artificial intelligence. As a co-creator and leader, he guided the framework from its initial development into a dominant force in the AI ecosystem ^[3].

Creation and Principles

PyTorch was developed as a successor to the Lua-based Torch framework, with the goal of creating a more user-friendly and Python-native deep learning library. Chintala's vision was to make AI research "joyful" and its tools "elegant," thereby lowering the barrier to entry for developers and researchers worldwide ^[3].

The framework's rapid adoption was driven by several key design principles:

• Python-First Integration: PyTorch was designed to feel native to Python, allowing developers to use standard Python control flow and debugging tools seamlessly ^[1].
• Dynamic Computation Graphs: It introduced a "define-by-run" approach, also known as eager execution. This allows the network's structure to change during runtime, offering greater flexibility for complex models and making the debugging process significantly easier compared to the static graphs used by competing frameworks at the time ^[5].
• Ease of Use: A simple and intuitive API was a core priority, which accelerated experimentation and innovation within the research community ^[1].

Impact and Adoption

Under Chintala's leadership, PyTorch quickly became the preferred framework in academia, used in over 80% of research papers published at major conferences like NeurIPS by 2023. Its adoption expanded across the industry, with estimates suggesting it is used by over 90% of AI practitioners and enterprises ^[2].

PyTorch now serves as a critical infrastructure layer for applications in generative AI, autonomous driving, and healthcare diagnostics. It is used by major technology companies including OpenAI, Google, Microsoft, Tesla, Cruise, and AstraZeneca, and powers features in consumer products from Instagram, TikTok, and Snapchat ^{[1] [2]}. In 2022, to ensure its long-term sustainability and community-driven governance, Meta transitioned PyTorch to the newly formed, vendor-neutral PyTorch Foundation, hosted by the Linux Foundation ^[4].

Open Source Contributions

Chintala is a vocal proponent of open-source software and research. His philosophy is reflected in his statements that "AI is delicious when it is accessible and open-source" and "Giving away technology and knowledge is one of the best ways to equal the playing field" ^{[4] [1]}. He has credited Meta's leadership, particularly Mark Zuckerberg and Mike Schroepfer, for supporting the view that "open-sourcing is fundamentally important and is a sound business strategy," which he noted was critical for the success of both FAIR and PyTorch ^[3].

Beyond PyTorch, his contributions include his foundational work on Torch7 and EBLearn. His GitHub profile showcases a history of involvement in various machine learning projects, including ganhacks, a repository of tips for training GANs, and convnet-benchmarks ^[5]. He has also emphasized the importance of community engagement, having personally answered thousands of user questions on forums to stay close to the "grassroots" of the developer ecosystem. Even after announcing his departure from Meta, Chintala confirmed his intention to remain an active member of the PyTorch community ^{[3] [1]}.

Future Plans

Following his departure from Meta, Chintala plans to pursue a new, smaller-scale venture. He described his next steps as building "Something small. Something new. Something I don't fully understand yet. Something uncomfortable," indicating an interest in early-stage or experimental projects ^[3]. He has also expressed interest in advising a select number of startups and exploring the world of venture capital. Chintala is already an active angel investor in several AI companies, including Anthropic, Runway, and Together.ai. He also continues to work on personal research projects in the field of home robotics, with the goal of building robots capable of automating household chores ^[1].