Sumit Agarwal
Software Engineer II at Microsoft
Bellevue, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolSumit Agarwal is a Software Engineer II at Microsoft in Bellevue with six years of experience focused on GPU compute programming and accelerating AI inference. He contributes to the widely used microsoft/onnxruntime project and the DirectML Execution Provider, enabling fp16 model tests across hardware and adding support for multiple Softmax/Hardmax/LogSoftmax operator versions to squeeze performance from diverse devices. Sumit pairs a strong academic foundation (MS in Computer Science from Stony Brook, 3.83/4.0; BE from BIT Mesra) with practical industry experience at Goldman Sachs and Uber, bringing production-grade rigor to ML runtimes. Motivated by the motto "Make AI run faster, faster and faster," he’s as comfortable tuning low-level numerical tolerances and device-specific validations as he is iterating on execution-provider features that matter at scale.
6 years of coding experience
3 years of employment as a software developer
Bachelor of Engineering (B.E.), Computer Science, 8.15/10.0 (Absolute Grading), Bachelor of Engineering (B.E.), Computer Science, 8.15/10.0 (Absolute Grading) at Birla Institute of TechnologyAISSCE PCM XII, Physics Chemistry Maths, 90.4%, AISSCE PCM XII, Physics Chemistry Maths, 90.4% at DELHI PUBLIC SCHOOL BOKAROMaster of Science - MS, Computer Science, 3.83/4.0, Master of Science - MS, Computer Science, 3.83/4.0 at Stony Brook University
English, Hindi, Marwari, Sanskrit
Github Skills (9)
machine-learning10
c-language10
deep-learning10
onnx10
c-programming-language10
hardware-acceleration9
tensorflow7
tensorflow27
pytorch7
Programming languages (3)
Github contributions (5)
microsoft/onnxruntime
Apr 2021 - Jan 2023
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
ML Engineer
ML EngineerContributions:313 reviews, 123 commits, 103 PRs in 1 year 9 months
Contributions summary:Sumit contributed to the testing and optimization of ONNX Runtime, specifically focusing on machine learning model inference and execution. They enabled and updated tests for the `fp16_inception_v1` model on various hardware platforms, modifying code related to test tolerance and device-specific configurations. The user also addressed issues with output tensor shape validation between ONNX inference and ONNX Runtime. In addition, the user worked on enhancing DML Execution Provider, including adding support for different versions of Softmax, Hardmax and LogSoftmax operator.
runtimetrainingtensorflowai-frameworkaccelerator
sumitsays/onnxruntime
Mar 2021 - May 2021
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Contributions:4 PRs, 15 pushes, 1 branch in 1 month
pytorchdeep-learningruntimemachine-learningonnx
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.