Avik Pal
ML Compiler Engineer at Waymo
Cambridge, Massachusetts, United States
Summary
🤩
Rockstar🎓
Top SchoolAvik Pal is an ML compiler engineer and PhD student at MIT with eight years of hands-on experience building high-performance compilers, scientific ML tools, and deep learning systems. His work spans deploying an MLIR-based compiler to thousands of TPUs and GPUs for ocean and hypersonics simulations, designing learned cost models integrated into XLA:TPU, and authoring numerics libraries with widespread usage (a nonlinear root-finding framework with 10k+ monthly unique downloads). He is a prolific open-source contributor across SciML and Flux ecosystems—implementing neural differential equation layers, depthwise convolutions, and custom autodiff adjoints—and has improved solver algorithms and GPU performance at scale. Comfortable at the intersection of theory and production, he combines deep knowledge of numerical methods, automatic differentiation, and system-level optimizations to make scientific ML both accurate and scalable. An often-overlooked thread through his work is rigorous testing and performance engineering: many contributions focus on correctness, reproducibility, and making advanced models practical in constrained compute environments.
8 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Massachusetts Institute of TechnologyIndian Institute of Technology KanpurCISCE, 10+2, Mathematics and Computer Science, CISCE, 10+2, Mathematics and Computer Science at National Gems H.S. School
Bengali, English, Hindi
Stackoverflow
36reputation
96reached
3answers
0questions
Github Skills (34)
algorithm10
algorithms10
pytorch10
ziggy10
convolutional-neural-networks10
ode10
cifar10010
python10
machine-learning10
gradient10
automatic-differentiation10
differential-equations10
zig10
deep-learning10
tensorflow10
Programming languages (13)
Github contributions (5)
torchgan/torchgan
Sep 2018 - Oct 2021
Research Framework for easy and efficient training of GANs based on Pytorch
Role in this project:
Backend Developer
Backend DeveloperContributions:5 releases, 90 commits, 102 PRs in 3 years 1 month
Contributions summary:Avik implemented project management features, including Travis CI, Codecov, and documentation support. The commits involve alterations to the project's licensing conditions and configuration files for documentation. Code modifications include adding a missing class and removing redundant code in the loss files.
pytorchpythondeep-learningtorchcomputer-vision
FluxML/Flux.jl
May 2018 - May 2019
Relax! Flux is the ML library that doesn't make you tensor
Role in this project:ML Engineer
ML EngineerContributions:100 commits, 12 PRs, 2 branches in 11 months
Contributions summary:Avik focused on implementing and integrating depthwise convolutional layers within the Flux.jl machine learning library. Their work included adding support for the DepthwiseConv layer, incorporating new constructors for this layer, and integrating it into the existing tracking and backpropagation system. The user also contributed to the test suite, adding tests to verify the correct functionality of the newly implemented depthwise convolution.
ml-librarythe-human-braindata-sciencedeep-learningneural-networks
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.