Fabian Pedregosa

Career

Fabian Pedregosa started in open source as a core contributor to the SymPy computer algebra system, then worked at INRIA on scikit-learn. Pedregosa went on to a PhD applying machine learning to brain imaging, followed by postdoctoral work on mathematical optimization at INRIA with Alexandre d'Aspremont and at UC Berkeley with Laurent El Ghaoui.¹

Pedregosa then joined Google Research, where the work focused on mathematical optimization until 2022 and on AI for mathematics afterwards. Pedregosa's homepage describes being part of the team behind the International Mathematical Olympiad results of 2024 and 2025, and current work on the Gemini Core team on math, code and related topics.¹

Open-source work

David Cournapeau started scikit-learn in 2007; Pedregosa took over the project in January 2010, made its first public release on 1 February 2010 and served as developer, maintainer and release manager until January 2012. Pedregosa is first author of the 2011 JMLR paper introducing the library and is listed among its emeritus maintainers.¹³²

Pedregosa started memory-profiler, a line-by-line memory profiler for Python, in 2012 and remains its lead developer. At Google, Pedregosa co-authored the implicit differentiation framework behind JAXopt, a differentiable optimization library for JAX that was discontinued in December 2023 in favor of Optax, and also contributes to Optax.¹⁴⁵

Work and contributions

• scikit-learn Open-source code

  Took over the project in 2010 and made its first public release, serving as developer, maintainer and release manager until 2012; first author of the 2011 JMLR paper and listed among the emeritus maintainers on the About page.¹²³⁶

• JAXopt (implicit differentiation) Published research

  Co-author of "Efficient and Modular Implicit Differentiation" (2021), which the JAXopt README names as the description of its implicit differentiation framework; Pedregosa's homepage lists JAXopt, discontinued in 2023 in favor of Optax, among the projects.⁴⁵¹

• memory-profiler Open-source code

  Started the line-by-line memory profiler for Python in 2012 and is described on the homepage as its lead developer and maintainer.¹⁷

Timeline

1. 2010Received a Google Summer of Code scholarship to improve the SymPy logic module.¹
2. Feb 1, 2010Made the first public release of scikit-learn after taking over the project in January 2010.¹
3. 2011Published "Scikit-learn, Machine Learning in Python" in JMLR as first author.³
4. 2012Started the memory-profiler project and handed off scikit-learn release management.¹
5. May 31, 2021Co-authored "Efficient and Modular Implicit Differentiation", the framework behind JAXopt.⁴
6. Dec 2023JAXopt was discontinued in favor of Optax.¹
7. 2024Was part of the Google team whose language model took part in the International Mathematical Olympiad.¹

Other documented roles

Roles the sources document that are not confirmed as current.

• Postdoc (with Laurent El Ghaoui), University of California, Berkeley¹

Honors and recognition

• EuroSciPy Best Poster Award (memory-profiler) · EuroSciPy 2012 conference, 2012¹

Where they work

• Google 121 people in the wiki

Related people

• Jeff DeanAlso at Google
• Michael CollinsAlso at Google
• Yuxin WuAlso at Google
• Martín AbadiAlso at Google
• Alex SmolaAlso at Google
• Michael AuliLanguage models · ML frameworks and libraries
• Razvan PascanuMachine-learning theory · ML frameworks and libraries
• Clément FarabetLanguage models · ML frameworks and libraries
• Beren MillidgeLanguage models · Machine-learning theory
• David TalbyLanguage models · ML frameworks and libraries