Ananth Subramaniam
Software Engineer at NVIDIA
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolAnanth Subramaniam is a software engineer with seven years of experience building scalable systems and machine learning tooling, now at NVIDIA after a long tenure at Meta. He combines backend and DevOps expertise with hands-on ML engineering, contributing to high-profile open-source projects like PyTorch Lightning and PyTorch TNT where he improved distributed training, logging, checkpointing, and fault tolerance. Comfortable across large-scale production environments, he has a track record of fixing critical issues in distributed training and adding practical utilities such as CUDA prefetchers and early-stopping logic. A UC Berkeley EECS graduate, he also has teaching experience as a reader for algorithms and ML courses, reflecting strong foundations in theory and practice. Based in the San Francisco Bay Area, he brings both research-adjacent rigor and production-focused pragmatism to ML infrastructure problems. Colleagues would note his tendency to tackle subtle edge cases in distributed systems that others often miss.
7 years of coding experience
9 years of employment as a software developer
Bachelor of Science (BS) Electrical Engineering and Computer Sciences, Bachelor of Science (BS) Electrical Engineering and Computer Sciences at University of California, BerkeleyThe Harker School
Github Skills (15)
neural-network10
pytorch10
machine-learning10
distributed-training10
pytorch-lightning10
deep-learning10
python10
pytest9
tensorboard9
automations8
automation8
data-pipeline8
data-pipelines8
automation-testing8
data-science7
Programming languages (3)
Github contributions (5)
Lightning-AI/pytorch-lightning
Jul 2020 - Nov 2022
Pretrain, finetune ANY AI model of ANY size on multiple GPUs, TPUs with zero code changes.
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:2487 reviews, 165 commits, 318 PRs in 2 years 3 months
Contributions summary:Ananth's contributions primarily focused on addressing code issues and improving the PyTorch Lightning library's functionality. The first set of commits addressed a critical issue related to environment variables used for rank zero casting in distributed training. The second contribution involved adding support for OmegaConf hparams to the Tensorboard logger, which included modifications to the logger code and related testing. Further contributions involved minor adjustments in testing as well as improvements to fault tolerant training and fixing issues for sharded data parallel training.
pythonheadachespytorch-modelsdata-sciencehandling
pytorch/tnt
Jul 2022 - Jan 2023
A lightweight library for PyTorch training tools and utilities
Role in this project:ML Engineer
ML EngineerContributions:23 reviews, 77 commits, 107 PRs in 6 months
Contributions summary:Ananth contributed to the development of training tools and utilities for PyTorch, a library for deep learning. Their work included refactoring code to improve organization, such as moving files under an existing `utils` directory and updating import paths. They also implemented version checks for PyTorch dependencies and added an early stopping checker, indicating involvement in model training workflows. Further contributions include a CUDA data prefetcher and a TorchSnapshot checkpoint saver.
pytorchpythondeep-learningreinforcement-learningmachine-learning
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.