Researcher

Zhihao Jia

Zhihao Jia is a computer scientist whose software systems help train and run neural networks efficiently. He earned a computer science bachelor's degree at Tsinghua University in 2013 and a PhD at Stanford in 2020, then joined Carnegie Mellon in 2021 after working as a research scientist at Facebook. He coauthored FlexFlow for distributed training, TASO for computation-graph optimization, SpecInfer for language-model serving and Mirage for GPU programs. He received a Sloan Research Fellowship in 2025. As of October 2026, he is an assistant professor in CMU's Computer Science Department and a co-founder of LithosAI, which accelerates AI agents on standard GPUs.

Now at
Carnegie Mellon University and LithosAI
Works
99
Citations
2,278
h-index
23
Badges

Publication numbers from OpenAlex, as of 27 October 2026.

Timeline

  1. 2012–2013

    Yao Class and MIT

    He studies as an exchange student at MIT in 2012 and completes a bachelor's degree in computer science at Tsinghua University in 2013. His undergraduate studies are in the Special Pilot CS Class, supervised by Andrew Yao.

  2. 2018

    FlexFlow for distributed training

    In 2018, he coauthors FlexFlow with Matei Zaharia and Alex Aiken. The deep learning system searches for fast ways to divide neural-network training across a parallel computer, using a simulator to predict how each strategy will perform.

  3. 2019

    TASO optimizes computation graphs

    He coauthors TASO in 2019, a system that automatically generates and verifies transformations of neural-network computation graphs. It searches among equivalent graphs to find an efficient way to run a model, and the work appears at the Symposium on Operating Systems Principles in October 2019.

  4. 2020

    A doctorate at Stanford

    In 2020, he earns his PhD from Stanford University's Computer Science Department, co-advised by Alex Aiken and Matei Zaharia.

  5. 2021

    Professor at Carnegie Mellon

    He joins Carnegie Mellon University's Computer Science Department as an assistant professor in 2021, after working as a research scientist at Facebook. Stanford awards him the Arthur Samuel PhD Best Thesis Award in 2021 for Automated Discovery of Machine Learning Optimizations.

  6. 2023

    SpecInfer for language models

    In 2023, he coauthors SpecInfer, a system that uses small models to predict possible outputs from a large language model. The large model checks a tree of candidate token sequences in parallel, reducing the computation needed to serve its outputs while preserving model quality.

  7. 2024

    Mirage for GPU programs

    He coauthors Mirage in 2024, a system that searches for optimized tensor programs for GPUs. It combines algebraic transformations, changes to execution schedules and custom kernels, and uses a probabilistic check to test whether an optimized program preserves the original computation.

  8. February 2025

    A Sloan Research Fellowship

    He receives a Sloan Research Fellowship in 2025. His work at CMU covers computer systems and compilers for machine learning; in 2025, he also co-directs the Catalyst group and is a member of the Parallel Data Lab.

  9. 2026

    Program chair for MLSys

    In 2026, he serves as a program co-chair of MLSys alongside Aakanksha Chowdhery, with affiliations at Carnegie Mellon University and LithosAI.

  10. As of October 2026

    Professor and LithosAI co-founder

    He is a co-founder of LithosAI with Dimitrios Skarlatos; the company accelerates AI agents on standard GPUs. As of October 2026, he is an assistant professor in Carnegie Mellon University's Computer Science Department, where he works on computer systems and machine learning.

Career

  1. Since 2021Carnegie Mellon UniversityAssistant Professor, Computer Science
  2. NowLithosAICo-Founder
  3. 2022–2024Tsinghua University
  4. MetaResearch Scientist

Education

  1. 2020Stanford UniversityPhD in computer science
  2. 2013Tsinghua UniversityBachelor’s degree in computer science
  3. Massachusetts Institute of TechnologyExchange Student

Sources and credits

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

  1. Zhihao Jia cs.cmu.edu. Checked 27 October 2026.
  2. Beyond Data and Model Parallelism for Deep Neural Networks arxiv.org. Checked 27 October 2026.
  3. GitHub - jiazhihao/TASO: The Tensor Algebra SuperOptimizer for Deep Learning github.com. Checked 27 October 2026.
  4. Recent Faculty Hires / 2021 - SCS Business Office - SCS Computing Facilities scsbusinessoffice.cs.cmu.edu. Checked 27 October 2026.
  5. Zhihao Jia | Carnegie Mellon University Computer Science Department csd.cmu.edu. Checked 27 October 2026.
  6. Research & Impact | Computer Science cs.stanford.edu. Checked 27 October 2026.
  7. SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verification arxiv.org. Checked 27 October 2026.
  8. Mirage: A Multi-Level Superoptimizer for Tensor Programs arxiv.org. Checked 27 October 2026.
  9. Faculty Members Named 2025 Sloan Research Fellows cs.cmu.edu. Checked 27 October 2026.
  10. Organizing Committee 2026 mlsys.org. Checked 27 October 2026.
  11. Portfolio - Unusual unusual.vc. Checked 27 October 2026.