Hongming Zheng
AI Engineer - Dynamo at Intel Corporation
California, United States
Summary
🤩
Rockstar🎓
Top SchoolHongming Zheng is an AI engineer and systems architect with over six years focused on large-scale distributed training and inference for LLMs and graph neural networks, currently advancing vLLM P/D disaggregation and heterogeneous serving across Intel platforms. He combines hands-on open-source contributions to PyTorch Geometric’s distributed training infrastructure with production-grade deployment experience using kubectl/helm and llm-d/vllm tooling. His background spans end-to-end IoT and smart manufacturing solutions, startup CTO experience that secured significant VC funding, and deep expertise in multi-node PyTorch, DeepSpeed, and quantized model performance tuning. At Intel he has led performance benchmarking and prefilling/decoding disaggregation for models like DeepSeek R1, Qwen and Mistral 7B, optimizing for TP/PP and low-memory strategies. Notably, he implemented distributed graph/feature stores and RPC-enabled partitioning for PyG, bridging research-grade GNNs to scalable training pipelines. Based in California, he pairs a PhD-level technical foundation with a practical record of moving complex AI systems from prototype to deployed service.
5 years of coding experience
17 years of employment as a software developer
Doctor of Philosophy (PhD), Doctor of Philosophy (PhD) at Southeast UniversityLarge Language Models for Business with Python (langchain/llamaindex/agent), Large Language Models for Business with Python (langchain/llamaindex/agent) at Stanford Continuing StudiesMulti AI Agent Systems with crewAI, Multi AI Agent Systems with crewAI at DeepLearning.AI
Github Skills (18)
pytorch10
ropc10
distributed-training10
python10
drpc10
graph10
graph-database10
rpc10
procedures10
graph-neural-network10
graph-datastructures10
deep-learning9
automations6
automation6
kubernetes4
Programming languages (1)
Github contributions (5)
pyg-team/pytorch_geometric
Jan 2023 - Feb 2024
Graph Neural Network Library for PyTorch
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:45 reviews, 19 PRs, 6 pushes in 1 year
Contributions summary:Hongming contributed significantly to the distributed training infrastructure of the PyTorch Geometric library. They implemented `LocalGraphStore` and `LocalFeatureStore` classes to handle distributed graph and feature data storage. Their work included adding helper initializations, partitioning algorithms, and RPC functionalities for remote feature lookups. The user also created examples and supporting scripts for partitioning and launching distributed training jobs.
pytorchgraph-convolutional-networksgeometric-deep-learningdeep-learningneural-graph
ZhengHongming888/pytorch_geometric
May 2023 - Feb 2024
Graph Neural Network Library for PyTorch
Contributions:5 PRs, 241 pushes, 35 branches in 9 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.