Yan Junjie is MiniMax’s founder, chairman, CEO and CTO, overseeing research strategy and company operations.[1]
Start with The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence →
Education
Doctorate in Artificial Intelligence
Completed Jul 2015
[1]Links
MiniMax managementFounder, Chief Executive Officer and Chief Technology Officer · MiniMax
Chairman and Executive Director · MiniMax
Career
Yan conducted postdoctoral research at Tsinghua and worked at SenseTime for more than six years before founding MiniMax.[1]
Contributions
Combine attention methods for long-context models
Yan is a named contributor to MiniMax-01. The report combines Lightning Attention with softmax attention and a mixture-of-experts architecture, explaining how that design supports long contexts in text and vision-language models.[2]
Train for agent work inside executable environments
The M2 report describes agent-generated trajectories grounded in workspaces and verifiable outputs, together with Forge, a reinforcement-learning system for long-running agent tasks. Yan appears in its alphabetic contributor list. The report evaluates a family of models for coding, search and workplace tasks; it represents the work of MiniMax’s larger research and engineering team.[3]
Tools, writing & talks
Writing, research & talks
Start here
A report on agent-driven data, executable workspaces and reinforcement-learning infrastructure.
Read the overview and agent-data pipeline before the Forge implementation details. Familiarity with language-model training and reinforcement learning helps; the examples connect model training to tasks with executable workspaces and checkable outputs.
Named team contributor[3]The team’s architecture and evaluation of long-context text and vision-language models.
Named team contributor[2]