Junyang Lin
Junyang Lin is an AI researcher who helped build Alibaba’s Qwen family of models. He joined Alibaba’s DAMO Academy in 2019 and led Qwen’s development from 2023, including work on Qwen2.5 and its variants. In 2025, he co-authored the Qwen3 technical report and a gated-attention paper that won a NeurIPS Best Paper award. As of October 2026, he is the founder of Pragmatik Labs, which researches agents for digital and physical worlds.
- Now at
- Pragmatik Labs
- Works
- 163
Publication numbers from OpenAlex, as of 15 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.
Timeline
-
2019
Joining Alibaba DAMO Academy
He joins Alibaba’s DAMO Academy in 2019 and works on language generation for Taobao search.
-
2023–2025
Leading Qwen at Alibaba
By December 2023, he leads Qwen at Alibaba. From 2023, he is responsible for developing, open-sourcing, and applying the Qwen series, including Qwen2.5, Qwen2.5-VL, Qwen2.5-Omni, and Qwen2.5-Coder.
-
2025
The Qwen3 technical report
He co-authors the Qwen3 technical report in 2025, which describes dense and mixture-of-experts models ranging from 0.6 to 235 billion parameters.
-
2025
A NeurIPS Best Paper
His co-authored paper “Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free” receives a NeurIPS 2025 Best Paper award.
-
March 2026
Stepping down from Qwen
He steps down from the Qwen project in March 2026 as Alibaba unveils its Qwen 3.5 open-weight small models.
-
Since August 2026
Founding Pragmatik Labs
In August 2026, he founds Pragmatik Labs in Shanghai to research next-generation agents across digital and physical worlds.
Sources and credits
Photo: We have no published photo of Junyang Lin yet. Send us a free-licensed photo.
- The man who built Qwen has a new lab, and the Shanghai government is an investor
- Qwen: Towards a Generalist Model
- PolyU Research Student Conference 2025 - Keynote II
- Qwen3 Technical Report
- Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
- Announcing the NeurIPS 2025 Best Paper Awards
- Alibaba’s Qwen tech lead steps down after major AI push
- Former Alibaba Qwen AI model leader founds new AI company with funding from venture giants