PeopleofSuperintelligence

Researcher

Edward J. Hu

Also known as Edward Hu

Edward Hu is an AI researcher whose work spans language-model adaptation, neural-network scaling and probabilistic reasoning. After studying computer science and linguistics in college and researching empirical NLP at Johns Hopkins, he worked as a researcher at Microsoft, where he coauthored the Maximal Update Parametrization (μP) in July 2021 and LoRA and μTransfer in 2022. At Mila in Montreal, he pursued PhD research under Yoshua Bengio, co-writing a March 2023 essay on generative flow networks and receiving an ICLR 2024 Outstanding Paper Honorable Mention for work on amortizing intractable inference in large language models. He also worked on o1 as a researcher at OpenAI, became a founding partner in a Woodside stealth AI company in 2025, and posted the introduction to his computer science PhD thesis, Building a Reasoning Machine, in April 2026. As of October 2026, he leads AI model training and research at Mercor.

Now at
Mercor

Timeline

  1. 2020

    An ICLR workshop prize

    After studying computer science and linguistics in college, he researches empirical NLP at Johns Hopkins, building systems to understand documents. In July 2020, his coauthored paper “Improved Image Wasserstein Attacks and Defenses” wins the Best Paper award at the ICLR Trustworthy ML Workshop.

  2. July 2021

    Maximal Update Parametrization at Microsoft

    Before graduate school, he works as a researcher at Microsoft. In July 2021, he and Greg Yang introduce the Maximal Update Parametrization (μP), a parametrization for infinitely wide neural networks that exhibit feature learning.

  3. 2022

    LoRA and μTransfer

    In March 2022, he, Greg Yang and Jianfeng Gao present μTransfer, a technique for hyperparameter tuning of enormous neural networks such as GPT-3. At ICLR in April 2022, he coauthors Low-Rank Adaptation (LoRA), which freezes pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture.

  4. 2023

    GFlowNets and PhD research at Mila

    In March 2023, he and Yoshua Bengio co-write “Scaling in the service of reasoning & model-based ML,” describing GFlowNets as a way to model compositional objects and build an inference machine for reasoning. By August 2023, he is an AI researcher at Mila in Montreal, pursuing his PhD under Bengio.

  5. 2024

    An ICLR 2024 Honorable Mention

    At ICLR 2024, his oral paper “Amortizing Intractable Inference in Large Language Models,” coauthored with Moksh Jain, Eric Elmoznino, Younesse Kaddar, Guillaume Lajoie, Yoshua Bengio and Nikolay Malkin, receives an Outstanding Paper Honorable Mention.

  6. 2025

    OpenAI o1 and a Woodside venture

    He works as a researcher on o1 at OpenAI, focuses his Mila PhD research on generative models and robust reasoning, and in 2025 is a founding partner in a stealth AI company in Woodside, California. The University of Montreal schedules his PhD defense, “Building a Reasoning Machine,” under research director Yoshua Bengio for November 2025.

  7. April 2026

    Building a Reasoning Machine

    He receives his PhD in computer science under the supervision of Yoshua Bengio. In April 2026, he posts the introduction to his PhD thesis, “Building a Reasoning Machine,” on reasoning and GFlowNets.

  8. As of October 2026

    AI model training at Mercor

    As of October 2026, he leads AI model training and research at Mercor.

Career

  1. NowMercorAI model training and research lead
  2. Johns Hopkins UniversityResearcher
  3. MicrosoftResearcher
  4. MilaAI Researcher

Education

  1. From 2025MilaPhD
  2. From 2025Université de MontréalPhD

From the biography.

  1. 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.
  2. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes OpenAI. 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.
  3. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  4. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Microsoft. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Edward J. Hu yet. Send us a free-licensed photo.

  1. Perspectives in AI: From LLMs to Reasoning with Edward Hu pear.vc. Checked 22 October 2026.
  2. Improved Image Wasserstein Attacks and Defenses microsoft.com. Checked 22 October 2026.
  3. About Edward Hu edwardjhu.com. Checked 22 October 2026.
  4. On infinitely wide neural networks that exhibit feature learning microsoft.com. Checked 22 October 2026.
  5. µTransfer: A technique for hyperparameter tuning of enormous neural networks microsoft.com. Checked 22 October 2026.
  6. LoRA: Low-Rank Adaptation of Large Language Models microsoft.com. Checked 22 October 2026.
  7. Scaling in the service of reasoning & model-based ML edwardjhu.com. Checked 22 October 2026.
  8. ICLR 2024 Awards iclr.cc. Checked 22 October 2026.
  9. Edward Hu — Deep Tech Week speaker deep-tech-week.com. Checked 22 October 2026.
  10. Soutenance de thèse - Edward Hu diro.umontreal.ca. Checked 22 October 2026.
  11. Building a Reasoning Machine - Ph.D. thesis edwardjhu.com. Checked 22 October 2026.