Jiawen Liu
Staff Research Scientist Meta Superintelligence Labs
United States
Summary
🤩
Rockstar🎓
Top SchoolJiawen Liu is a Staff Research Scientist at Meta Superintelligence Labs who leads efforts to scale and optimize GenAI inference and training systems, including contributions to the Llama 3 and Llama 4 families and work on efficient speculative decoding at production scale. With a PhD in Computer Science from UC Merced and three years of industry research experience, Jiawen bridges deep research and systems engineering to deliver high-performance model runtimes. Their open-source contributions to PyTorch’s Inductor compiler and FBGEMM GenAI demonstrate hands-on expertise in operator fusion, memory management, and GPU acceleration that produced measurable speedups for internal models. Prior internships at national labs and Tencent underscore a strong background in performance-focused research across academia and industry. Colleagues would describe Jiawen as a pragmatic problem-solver who pairs theoretical insight with low-level optimization to push large-scale multimodal AI into production.
3 years of coding experience
PhD, Computer Science, PhD, Computer Science at University of California, Merced
Github Skills (11)
pytorch10
machine-learning10
deeplearning-ai10
deep-learning10
gpu10
autograd9
python9
tensor9
neural-network8
c-language3
cprogramming-language3
Programming languages (2)
Github contributions (5)
pytorch/pytorch
Nov 2022 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:14 reviews, 10 commits, 12 PRs in 1 month
Contributions summary:Jiawen primarily focused on optimizing the performance of the Inductor compiler within the PyTorch framework. Their contributions involved building and integrating FX-based fusion strategies for linear, matmul, and permute operations, as well as shape padding for various matrix multiplications. These optimizations resulted in significant speedups for internal models. Additionally, they addressed and resolved issues related to memory management and graph manipulation within the Inductor compiler.
pythongpu-accelerationdeep-learninggpunumpy
jiawenliu64/FBGEMM
Nov 2022 - Mar 2025
FB (Facebook) + GEMM (General Matrix-Matrix Multiplication) - https://code.fb.com/ml-applications/fbgemm/
Contributions:62 pushes, 40 branches in 2 years 4 months
matrix-multiplicationlinear-algebrafacebookmultiplicationmatrix
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.