People Driving Open-Source AI Right Now

Thomas Wolf helped create the Transformers library, which made modern language models far easier for researchers and developers to use. The software became basic infrastructure for the open field.

Matei Zaharia created Apache Spark and MLflow before co-founding Databricks. Those open tools connect data engineering, model development and deployment across a huge range of organizations.

Liang Wenfeng built DeepSeek around efficient training and unusually open technical releases. Its models forced much larger laboratories to explain their costs and closed development choices.

Junyang Lin leads key work on Alibaba’s Qwen models, whose frequent open releases span sizes, languages and specialized tasks. That breadth has made Qwen a global developer platform.

Yann LeCun has used his scientific stature and position at Meta to argue consistently for open research and alternatives to language-model orthodoxy. His influence shapes both policy and technical debate.

Arthur Mensch turned Mistral into Europe’s most credible frontier-model company while releasing influential open-weight models. He balances commercial ambition with pressure to keep the ecosystem accessible.

Percy Liang founded Stanford’s Center for Research on Foundation Models and led HELM, an effort to evaluate models across many dimensions. Transparent measurement is essential to credible openness.

Clément Delangue turned Hugging Face into the main distribution and collaboration hub for open models, datasets and developer tools. That neutral ecosystem role reaches far beyond one company’s releases.

Yang Zhilin co-founded Moonshot AI and helped build the Kimi model family around long context and strong reasoning. The company adds a distinct Chinese voice to open-model competition.

Ion Stoica co-founded Anyscale and previously helped create Ray, the distributed-computing framework used to train and serve AI systems. His work makes open models practical at scale.