Taylan Bilal
Software Engineer at google
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolTaylan Bilal is a software engineer with nine years of experience specializing in machine learning infrastructure and TPU/XLA optimization, currently based in Zurich and working at Google. He brings a strong academic foundation—multiple master's degrees and a PhD in mathematics—and a history of applied data science roles from startups to Facebook. Taylan has made concrete open-source contributions to high-profile projects like pytorch/xla and fairseq, enabling broader model support and TPU compatibility for image classification and Wav2Vec2.0 training. He blends deep numerical expertise with pragmatic engineering, having improved test suites, fixed shape-related bugs, and optimized tensor initialization for production ML training on accelerators. Colleagues know him for translating research-grade models into robust, device-aware implementations that scale beyond single-GPU workflows.
9 years of coding experience
11 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Mathematics, 3.91, Doctor of Philosophy (Ph.D.), Mathematics, 3.91 at University of Southern CaliforniaMaster of Science (M.Sc.), Computational Science, 4.00, Master of Science (M.Sc.), Computational Science, 4.00 at Koc UniversityBS, Industrial Engineering, BS, Industrial Engineering at Galatasaray Üniversitesi
English, French, Turkish
Stackoverflow
36reputation
713reached
2answers
0questions
Github Skills (15)
xla10
pytorch10
machine-learning10
artificial-intelligence10
deep-learning10
python10
computer-vision9
nlp8
xgboost6
google-cloud-platform6
tpu6
tensorflow6
google-colaboratory6
compiler-compiler4
compiler4
Programming languages (5)
Github contributions (5)
pytorch/xla
Jun 2019 - Mar 2021
Enabling PyTorch on XLA Devices (e.g. Google TPU)
Role in this project:
ML Engineer
ML EngineerContributions:48 reviews, 59 commits, 78 PRs in 1 year 10 months
Contributions summary:Taylan contributed significantly to the `pytorch/xla` repository by enhancing the image classification test suite (`test/test_train_imagenet.py`). They added a `--model` flag to support various torchvision models, implemented model-specific configurations, and fixed bugs related to image dimensions. Furthermore, they integrated and tested multiple torchvision models, improving the test suite's coverage and usability for XLA devices. Their work involved adapting the existing training infrastructure to accommodate a broader range of models and ensuring their compatibility with the XLA platform.
pytorchxladeep-learningtpucompiler
facebookresearch/fairseq
Jul 2019 - Mar 2021
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:ML Engineer
ML EngineerContributions:1 review, 12 commits, 5 PRs in 1 year 8 months
Contributions summary:Taylan primarily focused on optimizing the performance of the `fairseq` library, specifically targeting TPU (Tensor Processing Units) compatibility. Their contributions included adding features necessary for progress bars to function with TPUs, modifying loss calculations for static shapes, and adjusting the code to enable Wav2Vec2.0 models to train effectively on TPUs. They also improved the code related to tensor initialization for improved efficiency on xla.
pytorchnlpsequencepythontransformer-architecture
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.