Sanket Purandare
Research Engineer at Meta
Greater Boston United States
Summary
🤩
Rockstar🎓
Top SchoolSanket Purandare is a research engineer and final-year PhD candidate at Harvard SEAS specializing in systems for deep learning, with nine years of experience optimizing training and inference through OS, compiler, and architecture techniques. He has bridged academic innovation and production at Meta, contributing to the PyTorch Distributed Stack and authoring components (like memory trackers and runtime estimators) that power scalable LLM training in open-source pytorch/pytorch. His work spans practical auto-parallel algorithms (Auto-SAC, Auto-FSDP), compiler-driven FSDP implementations, and novel algorithmic speedups for long convolution sequence models that delivered multi× end-to-end improvements. Prior projects include translating imperative PL-SQL to set-based SQL inside Microsoft SQL Server and ML-driven query plan selection at IISc, demonstrating a rare blend of systems research and production engineering. Based in Greater Boston, he pairs rigorous research output with hands-on engineering, and outside work he’s an amateur painter and a motorsports enthusiast—an indicator of creative problem-solving and attention to detail.
9 years of coding experience
6 years of employment as a software developer
Master of Technology (MTech), Computer Science, Master of Technology (MTech), Computer Science at Indian Institute of Science (IISc)Diploma, Computer Technology/Computer Systems Technology, Diploma, Computer Technology/Computer Systems Technology at Vivekanand Education Society's PolytechnicDoctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Harvard UniversityBachelor of Engineering (B.E.), Computer Engineering, Bachelor of Engineering (B.E.), Computer Engineering at Thadomal Shahani Engineering College
English, Hindi, Marathi
Github Skills (15)
performance-tuning10
pytorch10
machine-learning10
performance-monitor10
tensor10
performance-analysis10
performance-analytics10
performance-monitoring10
gpu10
python10
performance-measurement10
autograd9
backend9
neural-network9
deep-learning9
Programming languages (3)
Github contributions (5)
pytorch/pytorch
Oct 2022 - Feb 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:111 reviews, 16 PRs, 23 pushes in 2 years 5 months
Contributions summary:Sanket primarily contributed to enhancing the PyTorch module tracker, specifically focusing on its extension and functionality. Their work included implementing user-defined hooks for pre- and post-forward/backward operations, adding the `get_known_fqn` function, and optimizing multi-grad hook registration. Furthermore, the user implemented a memory tracker for FSDP modules and a runtime estimator. This work demonstrates a focus on core PyTorch functionality and performance optimization within the framework.
pythongpu-accelerationdeep-learninggpunumpy
sanketpurandare/pytorch
Apr 2023 - Mar 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Contributions:4 PRs, 181 pushes, 32 branches in 1 year 11 months
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.