Builders & open source
Weizhu Chen
Technical Fellow and Corporate Vice President, leading a modeling team, Microsoft
- Technical Fellow and Corporate Vice President, leading a modeling team, Microsoft AI1
Weizhu Chen is a Technical Fellow and Corporate Vice President at Microsoft AI, where Chen leads a modeling team working on large-scale model training with a recent focus on agentic and long-horizon reinforcement learning and synthetic data. Chen's Microsoft Research page lists the team's published work including LoRA, DeBERTa, Phi and Rho-1. Chen is senior author of DeBERTa (2020), LoRA (2021) and Samba (2024), and a co-author of the 2025 SambaY paper that describes the architecture of Phi-4-mini-flash-reasoning.12345
At a glance
- Current role
- Technical Fellow and Corporate Vice President, leading a modeling team
- Company
- Microsoft
- Sector
- Company
- Work listed
- 3
- Sources cited
- 7
- Profile reviewed
Role
Chen is a Technical Fellow and Corporate Vice President at Microsoft AI and leads a modeling team that trains large-scale models, including OpenAI and Microsoft AI models according to Chen's Microsoft Research page, with recent work on agentic and long-horizon reinforcement learning and synthetic data generation.1
Research
Chen is the senior author of DeBERTa (2020), a BERT variant with disentangled attention, and LoRA (2021), a method for adapting large language models with low-rank updates. Chen is also listed among the authors of the Phi-4-Mini technical report (2025).236
With Liliang Ren as first author, Chen co-authored Samba (2024), a hybrid state-space language model, and the 2025 SambaY paper, whose decoder-hybrid-decoder architecture is used in Phi-4-mini-flash-reasoning.457
Work and contributions
Phi-4-mini-flash-reasoning Open weights
Co-author of the SambaY decoder-hybrid-decoder paper, whose architecture Phi-4-mini-flash-reasoning adopts.57
LoRA Published research
Senior author of the LoRA paper, which adapts large language models by training low-rank update matrices while freezing the pretrained weights.3
DeBERTa Published research
Senior author of the DeBERTa paper on decoding-enhanced BERT with disentangled attention.2
Timeline
- Senior author of the DeBERTa paper on disentangled attention.2
- Senior author of the LoRA paper on low-rank adaptation of large language models.3
- Senior author of the Samba paper on hybrid state-space language models.4
- Co-authored the Phi-4-Mini technical report.6
- Co-authored the SambaY paper behind Phi-4-mini-flash-reasoning.5
Where they work
- Microsoft 75 people in the wiki
Related people
- Saurabh TiwaryAlso at Microsoft
- Furu WeiAlso at Microsoft
- David GrangierAlso at Microsoft
- Liliang RenAlso at Microsoft
- Karén SimonyanAlso at Microsoft
- Mika SenghaasReinforcement learning · Training systems
- Edward BeechingLanguage models · Reinforcement learning
- Susan ZhangTraining systems · Language models
- Sami JaghouarTraining systems · Reinforcement learning
- Jakub PachockiLanguage models · Reinforcement learning
Suggest a correction or removal
Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.
References
7 sources · profile reviewed Oct 22, 2026 · data updated Oct 22, 2026
- Weizhu Chen at Microsoft Research
- DeBERTa: Decoding-enhanced BERT with Disentangled Attention (arXiv 2006.03654)
- LoRA: Low-Rank Adaptation of Large Language Models (arXiv 2106.09685)
- Samba: Simple Hybrid State Space Models for Efficient Unlimited Context Language Modeling (arXiv 2406.07522)
- Decoder-Hybrid-Decoder Architecture for Efficient Reasoning with Long Generation (arXiv 2507.06607)
- Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs (arXiv 2503.01743)
- microsoft/Phi-4-mini-flash-reasoning model card