PeopleofSuperintelligence

Researcher

Shuaiwen Leon Song

Also known as Leon Song

Shuaiwen Leon Song is a computer scientist whose work includes GPU memory management, efficient language-model inference, and AI systems for scientific discovery. He earned a computer science doctorate at Virginia Tech and joined Pacific Northwest National Laboratory in 2013, then began academic work at the University of Sydney in 2019. His research included SuperNeurons for deep-network training and FP6-LLM for running language models with six-bit weights. At Microsoft, he helped launch DeepSpeed4Science in 2023 and served as its chief scientist before moving to Together AI. As of April 2026, he was Together AI’s vice president of research; his 2026 work included Aurora, which uses reinforcement learning to update speculative-decoding models during live serving.

Works
123
Citations
3,162
h-index
29
Badges

Publication numbers from OpenAlex, as of 27 October 2026.

Timeline

  1. 2013

    A doctorate and PNNL

    He earns a PhD in computer science at Virginia Tech in 2013. In 2013, he joins Pacific Northwest National Laboratory (PNNL).

  2. 2017

    An IEEE early career award

    In 2017, he receives the IEEE Technical Consortium on High Performance Computing Award for Excellence for Early Career Researchers. He works in PNNL’s High Performance Computing group, studying the interaction between computer hardware and software.

  3. January 2018

    SuperNeurons manages GPU memory

    He coauthors SuperNeurons in January 2018. The runtime manages GPU memory so that deep neural networks can be trained beyond a GPU’s DRAM capacity.

  4. 2019

    Joining Washington and Sydney

    In May 2019, he becomes an affiliated professor in the University of Washington’s Department of Electrical and Computer Engineering. He begins academic work at the University of Sydney in September 2019.

  5. 2020

    AI for cloud-edge virtual reality

    In 2020, Engineers Australia names him one of Australia’s most innovative engineers. His team develops an AI-based virtual-reality prototype that combines cloud and edge computing through hardware and software co-design.

  6. December 2022

    An IEEE scalable computing award

    He receives the IEEE mid-career award for scalable computing in December 2022.

  7. September 2023

    DeepSpeed4Science at Microsoft

    At Microsoft, he helps launch DeepSpeed4Science in September 2023 as a senior principal research manager. The initiative develops AI system technologies for scientific discovery, including protein-structure prediction and models of viral evolution.

  8. January 2024

    FP6-LLM for language-model inference

    In January 2024, he coauthors FP6-LLM, which combines six-bit model weights with TC-FPx GPU kernels for language-model inference. Its experiments run LLaMA-70B on a single GPU.

  9. By May 2024

    Vice president at Together AI

    By May 2024, he is vice president of research and frontier technologies at Together AI. Before the move, he serves as chief scientist for DeepSpeed4Science and manages Brainwave for Bing, a cloud-inference project at Microsoft. His earlier academic posts include SOAR associate professor at the University of Sydney.

  10. March 2026–April 2026

    Aurora at Together AI

    In March 2026, he coauthors Aurora at Together AI. The framework uses reinforcement learning to learn from live inference traces and update speculative-decoding models without interrupting serving. In April 2026, he is vice president of research at Together AI, where he leads research and development on large-scale inference systems.

Career

  1. NowTogether AIVice President of Research and Frontier Technologies
  2. NowUniversity of WashingtonAffiliate Assistant Professor, Electrical and Computer Engineering
  3. Until 2024MicrosoftChief Scientist for DeepSpeed4Science
  4. From 2023MicrosoftSenior Principal Research Manager
  5. From 2019University of WashingtonAffiliated Professor
  6. From 2019University of SydneySOAR Associate Professor
  7. From 2013Office of Science

Education

  1. 2013Virginia TechPhD in computer science

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Checked career sources: ee.washington.edu, ee.washington.edu.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Shuaiwen Leon Song yet. Send us a free-licensed photo.

  1. Shuaiwen Leon Song | UW Department of Electrical & Computer Engineering ee.washington.edu. Checked 27 October 2026.
  2. Song Named IEEE Computer Society Early Career Honoree | Staff Accomplishment | PNNL pnnl.gov. Checked 27 October 2026.
  3. [1801.04380] SuperNeurons: Dynamic GPU Memory Management for Training Deep Neural Networks arxiv.org. Checked 27 October 2026.
  4. Shuaiwen Leon Song, Ph.D. (he/him/his) shuaiwen-leon-song.github.io. Checked 27 October 2026.
  5. Engineers Australia - The University of Sydney sydney.edu.au. Checked 27 October 2026.
  6. Announcing the DeepSpeed4Science Initiative: Enabling large-scale scientific discovery through sophisticated AI system technologies - Microsoft Research microsoft.com. Checked 27 October 2026.
  7. [2401.14112] FP6-LLM: Efficiently Serving Large Language Models Through FP6-Centric Algorithm-System Co-Design arxiv.org. Checked 27 October 2026.
  8. Shuaiwen Leon Song (Together AI) – DeepSpeed4Science: Enabling System Support for Large Signature AI4Science Models at Scale – Department of Computer Science cs.uchicago.edu. Checked 27 October 2026.
  9. Aurora together.ai. Checked 27 October 2026.
  10. Events | DAPLab daplab.cs.columbia.edu. Checked 27 October 2026.