Keli Wen
Software Engineer at Google
Pudong, Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolKeli Wen is a software engineer with four years of experience building AI infrastructure, graph ML, and quantitative systems, currently working on AI agent orchestration and memory modules at Google while co-founding LLMQuant. Her background spans top-tier internships at AWS, Microsoft, and Amazon where she optimized distributed training and inference (DeepSpeed, vLLM, Triton) and contributed performance improvements to the popular DGL graph learning library. She has hands-on expertise in neighbor-sampling and GraphBolt optimizations for large-scale GNN training, and has applied graph data pipelines to LLM/GenAI tasks such as graph-based RAG and agent planning. Trained at Peking University and Wuhan University, she blends research-oriented rigor with product-minded engineering, and her GitHub ethos—think twice—reflects a careful, efficiency-first approach to ML infra.
4 years of coding experience
1 year of employment as a software developer
Bachelor's degree, Computer Science, Bachelor's degree, Computer Science at Wuhan UniversityMaster's degree, Computer Science, Master's degree, Computer Science at Peking University
Github Skills (14)
pytorch10
eval10
machine-learning10
deep-learning10
graph-neural-network10
python10
evaluation10
model-testing10
dgl10
data-structure9
algorithm9
cuda9
data-structures9
algorithms9
Programming languages (4)
Github contributions (5)
dmlc/dgl
Apr 2023 - Jan 2024
Python package built to ease deep learning on graph, on top of existing DL frameworks.
Role in this project:
ML Engineer
ML EngineerContributions:124 reviews, 59 PRs, 19 pushes in 8 months
Contributions summary:Keli contributed significantly to examples related to deep learning on graphs, specifically focusing on neighbor sampling techniques for GraphSAGE models. Their work included implementing and optimizing sampling strategies like NeighborSampler and MultiLayerFullNeighborSampler, crucial for efficient training and inference on large graphs. Furthermore, the user added examples and optimizations for link prediction tasks using GraphBolt, demonstrating a solid understanding of graph neural networks and efficient data loading pipelines. The user's contributions also include enhancements to the underlying GraphBolt framework, introducing new features such as random engines, and improvements to code related to sampling algorithms.
pytorchpythondeep-learningmachine-learninggraph-neural-networks
Mixture-AI/meta-llama-explain
Apr 2024 - Jul 2024
Explanation of the llama2 repo.
Contributions:48 reviews, 24 PRs, 44 pushes in 2 months
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.