Guilherme Leobas
Senior Software Engineer at Quansight
Minas Gerais, Brazil
Summary
🤩
Rockstar🎓
Top SchoolGuilherme Leobas is a Senior Software Engineer with 11 years of experience specializing in compilers, JITs, and high-performance back-end systems, currently working on PyTorch Compiler and Numba at Quansight. He brings deep expertise in LLVM-based optimizations and database internals from contributions to projects like Numba, HeavyDB, and PyTorch, where he implemented array/text type support, NumPy operations, and kernel-level improvements. His academic work on compiler optimizations (Ring Optimization) and practical experience with binary instrumentation inform a pragmatic approach to performance-sensitive code. An active open-source contributor, he has a track record of shipping tested, low-level features—such as SymInt support and array method handling—across major ML and data platforms. Based in Minas Gerais, Brazil, he pairs research rigor with production engineering and even trains for triathlons in his spare time, reflecting discipline and endurance beyond code.
11 years of coding experience
7 years of employment as a software developer
Computer Science, Computer Science at Technological University DublinMSc in Computer Science Compilers and Compilers Optimizations, MSc in Computer Science Compilers and Compilers Optimizations at Universidade Federal de Minas Gerais
English, Portuguese
Github Skills (31)
structures10
struct10
pytorch10
c-language10
python10
llvm10
databases10
machine-learning10
numpy10
tensorflow10
sql10
text-encoding10
command-line10
data-structures10
functional-programming10
Programming languages (13)
Github contributions (5)
numba/numba
Mar 2019 - Dec 2022
NumPy aware dynamic Python compiler using LLVM
Role in this project:
Back-end Developer & Library Contributor
Back-end Developer & Library ContributorContributions:489 reviews, 467 commits, 129 PRs in 3 years 9 months
Contributions summary:Guilherme's contributions focus on implementing and refining functionality within the Numba library, particularly related to NumPy array manipulation. Their work primarily involves adding support for the ``np.delete`` function, extending functionality for various NumPy array operations, including indexing, and improving the handling of exceptions. Furthermore, the user contributed to implementing several mathematical functions, such as ``np.cbrt``, and included the development and integration of associated test cases. The user's contributions involved modifying and extending Numba's core functionality for improved numeric array operations and performance, specifically the support of array methods in a dynamic setting.
cudapythonparallelnumpynumba
pytorch/pytorch
Aug 2020 - Apr 2021
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:308 reviews, 41 commits, 192 PRs in 8 months
Contributions summary:Guilherme contributed to the core functionality of the PyTorch library, focusing on machine learning-related operations. They implemented decompositions and added support for SymInt to the `torch.take_along_dim` function. Additionally, they made enhancements related to batch rule implementations for various functions. Further contributions included support for torch.cond in vmap and fixes related to prod and tensor subclass functionalities.
pythongpu-accelerationdeep-learninggpunumpy
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.