Pritam Damania
Member Of Technical Staff at OpenAI
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolPritam Damania is a seasoned software engineer with 12 years of experience specializing in distributed ML, distributed systems, databases, and filesystems, now contributing as a Member of Technical Staff at OpenAI. He has led core distributed training efforts at Meta—shaping PyTorch Distributed features like FSDP, RPC, distributed checkpointing, tensor/pipeline parallelism and DDP—and later built ML infrastructure at Tesla. His open-source work includes performance and reliability improvements to PyTorch (optimizing distributed linear algebra, fused matmuls, all-gather ops and DDP lifecycle APIs) and authoring advanced multi-GPU Transformer tutorials. Earlier roles include foundational contributions to HBase/HDFS at Facebook and early core database work at Yugabyte, demonstrating a consistent focus on high-throughput, low-latency systems at scale. Based in San Francisco, he blends deep systems engineering with practical ML infrastructure know-how, and often surfaces reliability and recovery improvements that aren’t obvious from feature lists alone.
12 years of coding experience
14 years of employment as a software developer
Bachelor's degree Computer Engineering, Bachelor's degree Computer Engineering at University of MumbaiMaster's degree Computer Science, Master's degree Computer Science at Stony Brook University
Hindi, English, Gujarati
Stackoverflow
21reputation
4kreached
0answers
1question
Github Skills (32)
algorithm10
algorithms10
pytorch10
c-language10
distributed-training10
python10
databases10
machine-learning10
distributed-systems10
pipelining10
data-structure10
transformer-models10
deep-learning10
gpu10
performance-optimization10
Programming languages (5)
Github contributions (5)
pytorch/pytorch
Oct 2020 - Dec 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Backend Developer
Backend DeveloperContributions:1040 reviews, 79 commits, 434 PRs in 2 years 2 months
Contributions summary:Pritam's contributions center around optimizing and improving the performance of distributed linear algebra operations within PyTorch. They focused on techniques like fusing matrix multiplications, utilizing optimized all-gather functions, and ensuring the correct use of data types (dtype) in these operations. Furthermore, the user addressed issues related to error handling and recovery within the distributed system, demonstrating a focus on both performance and reliability. They also implemented APIs to better manage DDP's resources and lifecycle.
pythongpu-accelerationdeep-learninggpunumpy
pytorch/tutorials
Sep 2020 - Apr 2021
PyTorch tutorials.
Role in this project:ML Engineer
ML EngineerContributions:14 reviews, 6 commits, 19 PRs in 7 months
Contributions summary:Pritam significantly contributed to the PyTorch tutorials repository, focusing on advanced topics like pipeline parallelism within the context of Transformer models. Their work involved implementing and refining a tutorial that demonstrates training Transformer models across multiple GPUs. They addressed batch size issues, optimized TensorPipe options, and corrected the codebase related to distributed training workflows within the tutorial.
deep-learningpytorchpytorch-tutorials
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.