Deng Cai
Research Scientist, Seed at ByteDance
Shanghai, China
Summary
👤
Senior🎓
Top SchoolDeng Cai is a research scientist with 11 years of experience advancing NLP and large language model systems, currently building frontier LLMs at ByteDance Seed in Shanghai. He completed a PhD at The Chinese University of Hong Kong and has a strong industrial research track record across Tencent AI Lab, AWS AI, Microsoft Research, Alibaba DAMO, and SUTD focused on retrieval-augmented generation, dialogue, machine translation, and AIGC. Deng has deep implementation experience in core ML tooling—contributing backend and Python/Cython improvements to the well-known DyNet neural network toolkit—bridging low-level math structures to practical CPU/GPU workflows. His work spans both fundamental research and production-ready systems, with repeated internships and collaborations that highlight cross-institutional impact. Fluent in turning research ideas into robust code, he blends theoretical insight with engineering rigor to push towards AGI-capable LLMs. An intriguing personal touch: he shares a playful GitHub bio hinting at a curiosity-driven, poetic side behind serious technical work.
11 years of coding experience
2 years of employment as a software developer
Master's degree Computer Science, Master's degree Computer Science at Shanghai Jiao Tong UniversityBachelor's degree Computer Science, Bachelor's degree Computer Science at Xiamen UniversityThe Chinese University of Hong Kong (CUHK)
Chinese, English
Github Skills (13)
tensorrt10
tensor10
tensorflow10
python10
cython10
operation10
cprogramming-language9
machine-learning9
c-language9
deep-learning9
api8
apidoc8
gpu-programming7
Programming languages (3)
Github contributions (5)
clab/dynet
Feb 2017 - Jul 2017
DyNet: The Dynamic Neural Network Toolkit
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:32 commits, 9 PRs, 41 comments in 5 months
Contributions summary:Deng contributed to the DyNet toolkit by focusing on the integration and enhancement of Python and Cython functionalities. They implemented and modified various batch manipulation operations for the library's expression graph. Moreover, the user added and refined methods for clipping parameters and set_value to optimize its usage across CPU and GPU platforms. The commits also involved improvements in the expression graph and the model initialization process, showing involvement with the underlying math structures for DyNet.
dynetdynamic-neural-networkdeep-learningneural-networksmachine-learning
jcyk/copyisallyouneed
Jul 2020 - Aug 2021
Code for our ACL2021 paper Neural Machine Translation with Monolingual Translation Memory
Contributions:47 commits, 7 pushes, 12 comments in 1 year
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.