Michael Carilli
Senior Applied Cryptography Engineer at Matter Labs
Albuquerque, New Mexico, United States
Summary
🤩
Rockstar🎓
Top SchoolMichael Carilli is a Senior Applied Cryptography Engineer with a decade of experience applying GPU, CUDA, and high-performance computing techniques to real-world systems. At Matter Labs he is building a GPU-accelerated zero-knowledge prover for zkSync to enable thousands of Ethereum transactions per second while preserving decentralization and security. Previously at NVIDIA he contributed to PyTorch and the widely used NVIDIA/apex mixed-precision and distributed-training toolkit, writing CUDA kernels and DistributedDataParallel improvements to speed and stabilize multi-GPU training. His PhD in physics and work porting Fortran CFD code to Kokkos-enabled C++ for GPUs and Xeon Phis drive a pragmatic ability to extract large speedups from complex numerical code. Based in Albuquerque, he combines deep numerical, systems, and cryptographic expertise to accelerate compute-heavy primitives like big-integer math, elliptic-curve operations, multi-scalar multiplication, and NTTs.
11 years of coding experience
7 years of employment as a software developer
BS, Physics, 4.0 GPA in major, BS, Physics, 4.0 GPA in major at University of Notre DameDoctor of Philosophy (PhD), Physics, 4.0 GPA, Doctor of Philosophy (PhD), Physics, 4.0 GPA at University of California, Santa Barbara
English, German, Spanish
Stackoverflow
381reputation
11kreached
0answers
14questions
Github Skills (15)
deepqlearning10
cuda10
optimizations10
pytorch10
deeplearning10
distributed-training10
multi-precision10
optimization10
machine-learning9
kernel9
caching6
cpython6
avx6
filesystems6
python-c-api6
Programming languages (5)
Github contributions (5)
NVIDIA/apex
May 2018 - Aug 2020
A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch
Role in this project:
ML Engineer
ML EngineerContributions:2 reviews, 94 commits, 165 PRs in 2 years 3 months
Contributions summary:Michael primarily contributed to the `apex` repository, a PyTorch extension for mixed precision and distributed training. Their contributions include modifications to the `DistributedDataParallel` module to improve parameter handling, and efficient bucketing and allreduce operations for optimized distributed training. Furthermore, the user made changes to the fused Adam optimizer and related CUDA kernels, indicating expertise in optimizing deep learning training.
pytorchraymixed-precisiondeep-learningtemporal-data
mcarilli/pytorch
Oct 2018 - Jun 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Contributions:537 pushes, 115 branches in 3 years 8 months
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.