Yang Zhilin
Founder and CEO, Kimi / Moonshot AI
Yang Zhilin is an AI researcher and the founder and chief executive of Kimi, the Moonshot AI model and product organization. His research record includes co-authorship of Transformer-XL and XLNet, followed by work associated with Kimi’s agentic models. He earned his doctorate at Carnegie Mellon in 2019. His profile connects published language-model research with leadership of an organization building models and practical AI tools.
Get AI Secret’s free daily briefing, including Yang Zhilin and the wider AI industry.
At a glance
Yang Zhilin is an AI researcher and the founder and chief executive of Kimi, the Moonshot AI model and product organization. His research record includes co-authorship of Transformer-XL and XLNet, followed by work associated with Kimi’s agentic models. He earned his doctorate at Carnegie Mellon in 2019. His profile connects published language-model research with leadership of an organization building models and practical AI tools.
- Current work
- Founder and CEO, Kimi / Moonshot AI
- Education
- Tsinghua University; Carnegie Mellon University
- Doctorate
- Language and Information Technology, 2019
- Research
- Transformer-XL; XLNet; Kimi models
Sources checked . Selected career background and milestones.
Career, in context.
Selected research, leadership and career milestones, newest first.
Kimi announces K3.1 Pro
Kimi announces K3.1 Pro and a Highspeed variant across its product family. The team schedules full model weights for November 4; that future release is not yet a completed milestone.
Source: Kimi announcement and research index ↗Kimi introduces K3
Kimi introduces its K3 model with native vision and a one-million-token context window. The team’s announcement is company context for Yang’s leadership, not sole personal authorship.
Source: Kimi announcement and research index ↗Kimi releases K2.6
The Kimi team releases K2.6, emphasizing coding, long-running task execution and agent swarms, with availability through its app, website, API and coding product.
Source: Kimi announcement and research index ↗Kimi introduces K2.5
Kimi introduces the native multimodal K2.5 model and an agent-swarm research preview. Its announcement describes visual coding and coordinated tool use by dynamically created agents.
Source: Kimi announcement and research index ↗Kimi introduces K2 Thinking
Kimi releases K2 Thinking as a model combining reasoning with tool use. The launch extends the team’s work on test-time scaling and agentic tasks.
Source: Kimi announcement and research index ↗Kimi releases K2
Kimi releases its K2 Base and Instruct models, describing a mixture-of-experts architecture with one trillion total parameters. The release is a collective research and engineering achievement.
Source: Kimi announcement and research index ↗Co-authors XLNet
Yang is the first listed author of the XLNet paper, which introduces a generalized autoregressive pretraining method and integrates ideas from the collaborative Transformer-XL work.
Source: XLNet research paper ↗Co-authors Transformer-XL
Yang is a credited co-author of Transformer-XL, a language-model architecture using segment-level recurrence and a positional encoding scheme to learn beyond fixed-length contexts.
Source: Transformer-XL research paper ↗Earns his Carnegie Mellon doctorate
Carnegie Mellon’s Language Technologies Institute lists Yang among its 2019 alumni with a PhD in Language and Information Technology, documenting his formal research training.
Source: Carnegie Mellon University ↗
Show earlier historyHide earlier history
The latest on Yang Zhilin
Newest firstReporting on this topic is coming soon. Explore our latest coverage.