Margaret Qian
Builder at Applied Compute
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMargaret Qian is a builder and engineering leader with a decade of experience delivering scalable AI and infrastructure systems from research labs to production at companies including Databricks, MosaicML, OctoML, and Facebook. She has led teams to productize model serving and app deployment platforms, and previously contributed as a core engineer to TVM’s Relay/QNN stack—working on quantization passes and bfloat16/ONNX interoperability for a widely used open-source deep learning compiler. Based in San Francisco, she blends hands-on ML compiler work with people leadership, moving projects from experimental prototypes to production-grade engines. Her background includes early research at Columbia’s Creative Machines Lab and internships at Google and Facebook, giving her a strong foundation in both theory and systems. Colleagues describe her as pragmatic and detail-oriented, able to navigate both low-level numeric challenges and high-level product tradeoffs.
10 years of coding experience
10 years of employment as a software developer
Bachelor's Degree, Computer Science, Bachelor's Degree, Computer Science at Columbia University
Chinese
Github Skills (12)
tvm10
relay-modern10
compiler10
machine-learning10
deeplearning-ai10
compiler-compiler10
deep-learning10
python10
relay10
onnx9
bfd8
tensorflow7
Programming languages (2)
Github contributions (5)
apache/tvm
Feb 2022 - May 2022
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
ML Engineer
ML EngineerContributions:14 reviews, 8 commits, 8 PRs in 3 months
Contributions summary:Margaret primarily contributed to the development of the TVM compiler stack, focusing on the "Relay" intermediate representation and its QNN (quantized neural network) capabilities. Their commits involved implementing and testing new passes for extracting and transforming fake quantized operations, specifically in the context of operators like adaptive average pooling, leaky ReLU, and mean. They also worked on improvements to existing QNN-related code, including fixing issues related to bfloat16 support, adding ONNX shape slicing capabilities, and making code formatting improvements.
metalvulkancompilertensoropencl
margaretqian/tvm
Jan 2022 - Jun 2022
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:28 pushes, 13 branches in 4 months
cpugpu-programminggpu-accelerationtvmdeep-learning
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.