PeopleofSuperintelligence

Researcher

Horace He

Also known as Chilli

Horace He is a machine learning systems researcher whose work spans PyTorch compilers and reproducible language model inference. His career includes Facebook and Google internships, PyTorch compiler work, and papers on torch.fx and FlexAttention. A move to Thinking Machines Lab in 2025 made him a founding engineer, and his research there addressed reproducible LLM inference. As of October 2026, his role at Thinking Machines Lab is researcher.

Timeline

  1. 2017–2019

    Facebook and Google internships

    His Facebook internships cover Nuclide for VSCode and dependency management in 2017, then PyTorch JIT and research with SysML in 2019. A Google internship in 2018 focuses on the SQLx compiler.

  2. 2020–2022

    PyTorch compilers and torch.fx

    PyTorch compiler work begins in 2020. The 2021 Torch.fx paper, coauthored by Horace He, introduces a Python library for capturing and transforming PyTorch programs; MLSys accepts the paper in 2022.

  3. 2024

    The PyTorch 2 paper

    At ASPLOS 2024, the PyTorch 2 paper and tutorial include him among the authors alongside Jason Ansel and Edward Yang.

  4. 2024

    FlexAttention kernels

    FlexAttention is a 2024 paper coauthored by Horace He that introduces a compiler-driven programming model for implementing many attention variants in PyTorch.

  5. March 2025

    Joining Thinking Machines

    In March 2025, his decision takes him from PyTorch, after about four years, to a founding engineer role at Thinking Machines Lab.

  6. September 2025

    Defeating LLM nondeterminism

    In September 2025, his research with Thinking Machines Lab collaborators shows that batch-invariant kernels make 1,000 completions of the same prompt identical.

  7. As of October 2026

    Researcher at Thinking Machines

    In October 2026, his role at Thinking Machines Lab is researcher.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Sources and credits

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

  1. Horace He horace.io. Checked 19 October 2026.
  2. Torch.fx: Practical Program Capture and Transformation for Deep Learning in Python arxiv.org. Checked 19 October 2026.
  3. PyTorch 2 paper and tutorial @ ASPLOS 2024 - PyTorch pytorch.org. Checked 19 October 2026.
  4. Flex Attention: A Programming Model for Generating Optimized Attention Kernels arxiv.org. Checked 19 October 2026.
  5. Why PyTorch is an amazing place to work... and Why I'm Joining Thinking Machines thonking.ai. Checked 19 October 2026.
  6. Defeating Nondeterminism in LLM Inference thinkingmachines.ai. Checked 19 October 2026.
  7. Thinking Machines launches open-source Inkling-Small model dataconomy.com. Checked 19 October 2026.