Yangqing Jia

Research

Jia co-authored DeCAF (2013), which showed that features from a deep convolutional network trained on one large image task transfer to other visual recognition problems. Jia was first author of the 2014 technical report on Caffe, a C++ framework with Python and MATLAB bindings for training and deploying deep models, which the Caffe site says Jia created during a PhD at UC Berkeley and which Berkeley AI Research and community contributors continued to develop.¹⁷⁴³⁸

As a research scientist at Google Brain, Jia worked on computer vision and deep learning and is a co-author of "Going Deeper with Convolutions" (2014), the paper introducing the GoogLeNet (Inception) architecture, which Jia's homepage lists as a co-created work. Jia was also part of the early design and implementation of TensorFlow and an author of its 2016 whitepaper.¹⁹⁵

Open-source frameworks

At Facebook, as a Director of Engineering, Jia initiated the company's general-purpose AI platform that served ranking, computer vision and language workloads. Jia's homepage credits Jia with creating Caffe2, co-creating ONNX, the Open Neural Network Exchange standard, and with co-leading PyTorch 1.0, which brought research and production uses of deep learning together; PyTorch's license retains a Caffe2 section that lists Jia's contributions.¹⁶¹⁰

Products and companies

Jia later served as a VP at Alibaba Group, as President of Alibaba Cloud's Computing Platform and head of the System AI Lab at DAMO Academy, before founding Lepton AI, an AI cloud company later acquired by NVIDIA, where Jia then worked as a VP on AI for system software and cloud services. In January 2024 Lepton AI open-sourced Lepton Search, a demo conversational search engine under the Apache-2.0 license. Jia now leads Intent Lab as founder and CEO, a company focused on turning user intent into production systems.¹¹¹²

Work and contributions

• PyTorch (1.0 and Caffe2) Open-source code

  Co-lead of PyTorch 1.0 and creator of Caffe2, according to Jia's homepage. PyTorch's BSD-style license file has a "From Caffe2" section that lists contributions by Yangqing Jia.¹⁶

• ONNX (Open Neural Network Exchange) Open-source code

  Co-creator of the ONNX standard, per Jia's homepage, which lists the work under Jia's role at Facebook; the ONNX repository is licensed under Apache 2.0.¹¹⁰

• Caffe Open-source code

  Created Caffe during a PhD at UC Berkeley, per Caffe's official site, and is first author of the Caffe paper (2014), which describes a BSD-licensed C++ library with Python and MATLAB bindings for training and deploying deep models; the repository is released under the BSD 2-Clause license.³⁴⁸

• TensorFlow Open-source code

  Worked on TensorFlow's early design and implementation at Google Brain, per Jia's homepage, and is a co-author of the TensorFlow whitepaper (2016); TensorFlow is licensed under Apache 2.0.¹⁵¹²

Timeline

1. Oct 2013Co-authored DeCAF, on reusing deep convolutional activation features for generic visual recognition.⁷
2. Jun 2014Published the Caffe technical report as first author, describing the deep learning framework.⁴
3. Sep 2014Co-authored "Going Deeper with Convolutions", introducing the GoogLeNet (Inception) architecture.⁹
4. Mar 2016Co-authored the TensorFlow whitepaper on large-scale machine learning on heterogeneous systems.⁵
5. Jan 2024Lepton AI, which Jia founded, open-sourced the Lepton Search conversational search demo.¹¹¹

Other documented roles

Roles the sources document that are not confirmed as current.

• VP, building AI for system software and cloud services for AI, NVIDIA¹
• Founder and CEO (Lepton AI was acquired by NVIDIA), Lepton AI¹
• VP; President of Alibaba Cloud's Computing Platform and Head of the System AI Lab at DAMO Academy, Alibaba Group¹
• Director of Engineering; built Facebook's general-purpose AI platform and co-led PyTorch 1.0, Facebook¹
• Research Scientist, Google Brain¹⁵
• PhD student, EECS (created Caffe during the PhD), University of California, Berkeley¹³

Where they work

• Intent Lab 1 person in the wiki

Related people

• Tong HeML frameworks and libraries · Computer vision
• Francisco MassaML frameworks and libraries · Computer vision
• Yuxin WuComputer vision · ML frameworks and libraries
• Barret ZophTraining systems · Computer vision
• Adam PaszkeML frameworks and libraries · Training systems
• William FalconML frameworks and libraries · Training systems
• Martín AbadiML frameworks and libraries · Training systems
• James BradburyML frameworks and libraries · Training systems
• Edward YangML frameworks and libraries · Training systems
• Luca AntigaML frameworks and libraries · Training systems