Yida Wang
Software Engineer at Google
Palo Alto, California, United States
Summary
🤩
Rockstar🎓
Top SchoolYida Wang is a Principal Scientist based in Palo Alto with 13 years of experience building ML systems, compilers, and performance-oriented infrastructure for large foundation models at AWS and previously Intel. He leads research and engineering efforts that enabled core Amazon products—from Trainium/Inferentia accelerators to Bedrock, Titan LLMs, and SageMaker compiler and distributed training services—publishing peer-reviewed work at top systems and ML venues. Technically fluent in low-level optimization and scalable parallelism, he contributes to notable open-source projects like TVM and Brainiak, improving CPU parallelization, MAC counting, and large-scale neuroimaging workflows. Trained in CS and neuroscience (PhD, Princeton), he combines rigorous research with production-grade engineering to make demanding ML workloads efficient and reliable on heterogeneous hardware. An under-the-radar strength is his track record of turning research prototypes into productized components across hardware, compiler, and service layers.
12 years of coding experience
2 years of employment as a software developer
Bachelor of Science - BS, Computer Science, Senior, Bachelor of Science - BS, Computer Science, Senior at Massachusetts Institute of Technology
Chinese
Github Skills (20)
c-language10
performance-monitor10
python10
scikit10
machine-learning10
numpy10
deeplearning-ai10
compiler-compiler10
performance-analysis10
deep-learning10
scikit-learn10
compiler10
cprogramming-language10
blas9
tensor9
Programming languages (4)
Github contributions (5)
brainiak/brainiak
Jul 2016 - Apr 2018
Brain Imaging Analysis Kit
Role in this project:
Back-end Developer / ML Engineer
Back-end Developer / ML EngineerContributions:25 commits, 27 PRs, 190 comments in 1 year 9 months
Contributions summary:Yida made significant contributions to the Brain Imaging Analysis Kit, focusing on the development and enhancement of the FCMA (Full Correlation Matrix Analysis) module. They implemented CMake support for installation, initialized FCMA functionality, and added Fisher transformation. Moreover, they introduced and refined various aspects of the classifier module, including support for precomputed kernels, partial similarity matrix computation, and the integration of multiprocessing for cross-validation. These enhancements improved performance and functionality related to voxel selection and classification.
brainconnectivityneuroimagingneurosciencepython
apache/tvm
Jan 2018 - Feb 2020
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:ML Engineer
ML EngineerContributions:19 commits, 18 PRs, 119 comments in 2 years 1 month
Contributions summary:Yida contributed to optimizing and improving the performance of deep learning models within the TVM compiler stack. Their work included implementing and modifying schedules for improved CPU parallelization, specifically addressing scalability issues in multi-threaded environments. The user also added and modified device properties, demonstrating involvement in hardware-specific optimizations. Furthermore, they added and modified MAC count for various operations, demonstrating deep knowledge of performance analysis of deep learning models.
metalvulkancompilertensoropencl
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.