Nicolas Macchioni
Software Engineer at Meta
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolNicolas Macchioni is a software engineer with seven years of experience building high-performance ML and systems software, currently contributing to PyTorch at Meta out of Menlo Park. A Carnegie Mellon CS graduate and former Facebook intern, he has shipped CUDA kernels and INT8 quantized GEMM support that delivered substantial throughput gains for large-model training on A100s. His open-source contributions to the widely-used pytorch/pytorch codebase span transformer masked softmax, multi-head attention fastpaths, and InDuctor compilation fixes—work that improves both correctness and runtime efficiency. Past internships at Lawrence Livermore National Laboratory and leadership on backend systems demonstrate a strong foundation in C/C++, Python, and large-scale simulation tooling. Outside engineering, he pairs a technical lens with a minor in photography, bringing an observant, detail-oriented approach to both code and composition.
7 years of coding experience
1 year of employment as a software developer
High School Diploma, General Education, Graduate, High School Diploma, General Education, Graduate at Granada High SchoolBachelor of Science - BS, Computer Science, Bachelor of Science - BS, Computer Science at Carnegie Mellon University
English
Stackoverflow
11reputation
4kreached
0answers
1question
Github Skills (13)
neural-network10
pytorch10
machine-learning10
deeplearning-ai10
tensor10
deep-learning10
python10
autograd9
gpu9
transformer9
numpy8
bytecode6
reverse-engineering6
Programming languages (5)
Github contributions (5)
pytorch/pytorch
Aug 2022 - Aug 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:125 reviews, 2 commits, 134 PRs in 1 day
Contributions summary:Nicolas primarily contributed to the PyTorch codebase, specifically focusing on improvements and fixes related to masked softmax and multi-head attention mechanisms within the transformer architecture. They added a mask identifier for the transformer fastpath to address incorrect interpretations of masks and supported src_mask in the masked_softmax CPU path. Additionally, the user's work involved debugging and fixing issues in the Inductor compilation framework, including bug fixes and improvements to autotuning. The contributions were focused on improving the performance, correctness, and efficiency of PyTorch's core features.
pythongpu-accelerationdeep-learninggpunumpy
nmacchioni/pytorch
Jul 2022 - Jul 2024
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Contributions:115 pushes, 35 branches in 2 years
pythongpu-accelerationdeep-learninggpuacceleration
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.