Johannes Hagemann

Career

Hagemann joined Aleph Alpha in early 2022 to help build a distributed training framework. At the end of 2023, Hagemann and Vincent Weisser committed to building Prime Intellect, which Hagemann describes as a stack for open AI, from agentic models to the infrastructure used to train, fine-tune and deploy them. Hagemann's GitHub profile lists the role of co-founder and CTO.¹²

Research

Hagemann co-authored OpenDiLoCo (2024), an open-source framework for distributed training with low communication between sites, and the INTELLECT-1 technical report (2024) on a model trained across distributed infrastructure. Later reports include INTELLECT-2 (2025), a reasoning model trained through globally decentralized reinforcement learning, and INTELLECT-3 (2025).⁶³⁴⁵

Hagemann is also a co-author of TOPLOC (2025), a locality-sensitive hashing scheme for verifiable inference, and of METAGENE-1 (2025), a metagenomic foundation model for pandemic monitoring.⁷⁸

Work and contributions

• INTELLECT model series Open weights

  Co-author of the INTELLECT-1 (2024) and INTELLECT-2 (2025) technical reports, which describe Prime Intellect's models trained on globally distributed compute.³⁴

• INTELLECT-3 Open weights

  Co-author of the INTELLECT-3 technical report (2025), which describes a mixture-of-experts model post-trained with large-scale reinforcement learning.⁵

• OpenDiLoCo Published research

  Co-author, with Sami Jaghouar and Jack Min Ong, of OpenDiLoCo, an open-source framework for globally distributed low-communication training.⁶

Timeline

1. 2022Joined Aleph Alpha to help build a distributed training framework.¹
2. 2023Committed with Vincent Weisser to building Prime Intellect at the end of the year.¹
3. Jul 2024Co-authored OpenDiLoCo, a framework for globally distributed low-communication training.⁶
4. Dec 2024Co-authored the INTELLECT-1 technical report.³
5. Jan 2025Co-authored the TOPLOC and METAGENE-1 papers.⁷⁸
6. May 2025Co-authored the INTELLECT-2 technical report on decentralized reinforcement learning.⁴
7. Dec 2025Co-authored the INTELLECT-3 technical report.⁵

Other documented roles

Roles the sources document that are not confirmed as current.

• Worked on a distributed training framework, from early 2022, Aleph Alpha¹

Where they work

• Prime Intellect 3 people in the wiki

Related people

• Mika SenghaasAlso at Prime Intellect
• Sami JaghouarAlso at Prime Intellect
• Edward BeechingLanguage models · Reinforcement learning
• Jakub PachockiLanguage models · Reinforcement learning
• Weizhu ChenLanguage models · Training systems
• Susan ZhangTraining systems · Language models
• Colin RaffelLanguage models · Training systems
• Shixiang Shane GuLanguage models · Reinforcement learning
• Naveen RaoTraining systems · Language models
• Liliang RenLanguage models · Training systems