tsurumeso
Software Engineer
Japan
Summary
🤩
Rockstartsurumeso is a Software Engineer based in Japan with 9 years of experience focused on machine learning and performant numerical computing. They have hands-on expertise building and training deep neural networks—most notably a UNet-based vocal remover—and implementing GPU-accelerated array operations in widely used open-source libraries like Chainer and CuPy. Their contributions span model architecture, dataset engineering and augmentation, and robust test automation, reflecting a balance of research-minded ML work and production-quality backend engineering. Notably, they implemented and hardened core array functions (flip, rot90) in CuPy/Chainer with comprehensive tests and documentation, improving GPU parity with NumPy. Colleagues can expect pragmatic code, attention to numerical correctness, and a track record of improving both ML models and the libraries that run them.
9 years of coding experience
Github Skills (26)
enet10
pytorch10
python10
testing10
data-manipulation10
machine-learning10
audio-processing10
manipulation10
numpy10
deep-learning10
gpu10
cupy10
cuda10
spectrogram9
linear-algebra9
Programming languages (5)
Github contributions (5)
tsurumeso/vocal-remover
Jun 2019 - Sep 2022
Vocal Remover using Deep Neural Networks
Role in this project:
ML Engineer
ML EngineerContributions:24 releases, 4 reviews, 155 commits in 3 years 2 months
Contributions summary:Tsurumeso's commits primarily focused on implementing and training a vocal remover model using deep neural networks. They developed a UNet-based architecture for vocal separation, utilizing PyTorch and Chainer frameworks. The user was also responsible for data handling, including dataset creation and augmentation techniques, and integrated various spectral analysis tools. The code includes the design and implementation of a multi-band U-Net model, further indicating expertise in model architecture.
pytorchremoverspectrogramdeep-learningvocal-separation
cupy/cupy
Jul 2017 - Aug 2018
NumPy & SciPy for GPU
Role in this project:Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:34 commits, 5 PRs, 10 comments in 1 year 1 month
Contributions summary:Tsurumeso implemented the `flip` and `rot90` functions within the CuPy library, mirroring the functionality of NumPy's equivalent functions. They fixed an axes check in the `rot90` function, ensuring correct behavior. In addition to function implementations, the user added comprehensive tests for the `flip` and `rot90` functions, including various test cases to cover different scenarios and data types. The user also refactored the `flip` and `rot90` functions and added documentation.
cudapythoncusolvergpunumpy
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.