PeopleofSuperintelligence

Researcher

Heewoo Jun

Heewoo Jun is an AI researcher whose work spans speech recognition, music generation, image pretraining, code models and 3D synthesis. He studied at the University of Toronto from 2008 to 2013 and was an MS student at Stanford University from 2013 to 2015 before working as a researcher at Baidu from 2015 to 2019, where he coauthored 2017 papers on GAN-based speech recognition and deep learning scaling. In 2020, he coauthored Jukebox for raw audio generation and Generative Pretraining From Pixels, which received an ICML Outstanding Paper honorable mention. In 2021, he coauthored papers introducing the Codex code model and the GSM8K math word problem dataset, followed by the 3D generative models Point-E in 2022 and Shap-E in 2023 and a 2024 US patent with Mohammad Bavarian for language-model text insertion. He is a researcher at OpenAI, where he has worked since 2019.

Now at
OpenAI
Badges

Timeline

  1. 2008–2015

    Toronto and Stanford study

    He studies as an undergraduate at the University of Toronto from 2008 to 2013, then as an MS student at Stanford University from 2013 to 2015.

  2. 2015–2019

    Speech and scaling research at Baidu

    From 2015 to 2019, he works as a researcher at Baidu. In 2017, he coauthors Robust Speech Recognition Using Generative Adversarial Networks, an end-to-end framework that uses a GAN objective to enable robust speech recognition, and Deep Learning Scaling is Predictable, Empirically, a study of how generalization error and model size grow with training sets.

  3. 2020

    Jukebox for raw audio

    He coauthors Jukebox in 2020, a model that generates music with singing in the raw audio domain. It uses a multi-scale VQ-VAE to compress raw audio into discrete codes and autoregressive Transformers to model them.

  4. 2020

    Image pretraining from pixels

    In 2020, Jun coauthors Generative Pretraining From Pixels, which trains a sequence Transformer to autoregressively predict pixels without incorporating knowledge of the 2D input structure. ICML gives the paper an Outstanding Paper honorable mention in 2020.

  5. 2021

    Codex and the GSM8K benchmark

    He coauthors Evaluating Large Language Models Trained on Code in 2021, introducing Codex, a GPT language model fine-tuned on GitHub code that powers GitHub Copilot. In 2021, he also coauthors Training Verifiers to Solve Math Word Problems, which introduces the GSM8K dataset of 8.5K grade school math word problems.

  6. 2022–2023

    Two generative models for 3D

    In 2022, he coauthors Point-E, which first generates a synthetic view with a text-to-image diffusion model and then produces a 3D point cloud with a second diffusion model. In 2023, he coauthors Shap-E, a conditional generative model for 3D assets.

  7. 2024

    Patent for text insertion

    Jun and Mohammad Bavarian receive US patent 11,886,826 in 2024 for systems that use a language model to automatically generate and insert text.

  8. Since 2019

    Research at OpenAI

    He is a researcher at OpenAI, where he has worked since 2019.

Career

  1. Since 2019OpenAIResearcher
  2. 2015–2019BaiduResearcher

Education

  1. 2015Stanford UniversityMS
  2. 2013University of TorontoUndergraduate student

From the biography.

  1. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Heewoo Jun. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  2. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Heewoo Jun. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  3. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes OpenAI. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  4. The median h-index is 48 at universities and 14 at frontier AI labsIncludes OpenAI. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.

Colleagues

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

Sources and credits

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

  1. Heewoo Jun (OpenAI): deep learning optimization — Lacuna lacuna.tiptreesystems.com. Checked 22 October 2026.
  2. [1711.01567] Robust Speech Recognition Using Generative Adversarial Networks arxiv.org. Checked 22 October 2026.
  3. [1712.00409] Deep Learning Scaling is Predictable, Empirically arxiv.org. Checked 22 October 2026.
  4. [2005.00341] Jukebox: A Generative Model for Music arxiv.org. Checked 22 October 2026.
  5. Generative Pretraining From Pixels proceedings.mlr.press. Checked 22 October 2026.
  6. Awards: Test of Time and Best Papers icml.cc. Checked 22 October 2026.
  7. [2107.03374] Evaluating Large Language Models Trained on Code arxiv.org. Checked 22 October 2026.
  8. [2110.14168] Training Verifiers to Solve Math Word Problems arxiv.org. Checked 22 October 2026.
  9. [2212.08751] Point-E: A System for Generating 3D Point Clouds from Complex Prompts arxiv.org. Checked 22 October 2026.
  10. [2305.02463] Shap-E: Generating Conditional 3D Implicit Functions arxiv.org. Checked 22 October 2026.
  11. US11886826B1 - Systems and methods for language model-based text insertion patents.google.com. Checked 22 October 2026.