PeopleofSuperintelligence

Builder

Zhifeng Chen

Zhifeng Chen is an engineer of large-scale machine learning systems, from TensorFlow at Google Brain to Apple's foundation models group. A Fudan University alumnus, he earned a master's degree at Princeton, completed a computer-science doctorate at the University of Illinois Urbana-Champaign in 2005, and joined Google in 2005. In 2016 he co-authored the OSDI paper on TensorFlow, a Google Brain dataflow system that Google released as open source, and a paper on GNMT, Google's neural machine translation system. Later work includes GPipe, a 2019 library for pipeline parallelism, and GShard, which in 2020 scaled a mixture-of-experts translation model beyond 600 billion parameters on 2048 TPU v3 accelerators. Since July 2025 he oversees Apple's foundation models group, which reports to Daphne Luong.

Now at
Apple
Works
66

Publication numbers from OpenAlex, as of 17 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2004

    Second-level buffer caches

    With Yuanyuan Zhou and Kai Li, he publishes on second-level buffer cache management in June 2004. The paper introduces Multi-Queue, a local algorithm for those caches.

  2. 2005

    A doctorate and Google

    A Fudan University alumnus, he earns a master's degree at Princeton and a doctorate at the University of Illinois Urbana-Champaign. In 2005 the Illinois PhD, in computer science, is on optimization of data accesses for database applications, with Yuanyuan Zhou as doctoral committee chair. He joins Google in 2005.

  3. 2016

    TensorFlow at Google Brain

    In 2016 he co-authors the OSDI paper on TensorFlow, a Google Brain system that uses dataflow graphs and maps computation across machines and across CPUs, GPUs, and Tensor Processing Units. The paper reports TensorFlow's release as an open-source project. He works as a Distinguished Software Engineer at Google Brain on large-scale computer systems and machine learning applications.

  4. 2016

    Google's translation system

    In 2016 he co-authors the paper on GNMT, Google's neural machine translation system. The model is a deep LSTM network with 8 encoder and 8 decoder layers using attention and residual connections.

  5. 2019

    The GPipe training library

    In March 2019 Google introduces GPipe, an open-source library that uses pipeline parallelism to train networks made of sequential layers. He co-authors the NeurIPS 2019 paper on the library.

  6. 2020

    GShard for giant models

    In 2020 he co-authors GShard, a system for conditional computation and automatic sharding. It scales a multilingual translation Transformer with a sparsely gated mixture of experts beyond 600 billion parameters, and the model trains on 2048 TPU v3 accelerators in 4 days.

  7. Since July 2025

    Apple foundation models

    In July 2025 he oversees Apple's foundation models group. The group reports to Daphne Luong, head of AI research at Apple.

Sources and credits

Photo: We have no published photo of Zhifeng Chen yet. Send us a free-licensed photo.

  1. Second-level buffer cache management collaborate.princeton.edu. Checked 16 October 2026.
  2. Trapped by silence: How Apple lost its top AI talent to Meta kr-asia.com. Checked 16 October 2026.
  3. Optimization of Data Accesses for Database Applications ideals.illinois.edu. Checked 16 October 2026.
  4. TensorFlow: A System for Large-Scale Machine Learning usenix.org. Checked 16 October 2026.
  5. TensorFlow: A system for large-scale machine learning research.google. Checked 16 October 2026.
  6. Zhifeng Chen zhifengchen.me. Checked 16 October 2026.
  7. Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation research.google. Checked 16 October 2026.
  8. Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation arxiv.org. Checked 16 October 2026.
  9. GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism research.google. Checked 16 October 2026.
  10. Introducing GPipe, an Open Source Library for Efficiently Training Large-scale Neural Network Models research.google. Checked 16 October 2026.
  11. GShard: Scaling Giant Models With Conditional Computation and Automatic Sharding research.google. Checked 16 October 2026.
  12. GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding arxiv.org. Checked 16 October 2026.
  13. Apple loses fourth AI researcher in a month to Meta's superintelligence team latimes.com. Checked 16 October 2026.