Hariharan Seshadri
Senior Software Engineer - AI Frameworks at Microsoft
Sunnyvale, California, United States
Summary
🤩
Rockstar🎓
Top SchoolHariharan Seshadri is a Senior Software Engineer at Microsoft with 11 years of experience building and optimizing AI frameworks, particularly around the ONNX standard and ONNX Runtime. He combines deep backend and ML engineering skills—implementing datatype and operator support across CPU and CUDA providers, shape inference, and execution planning—with production-minded performance work such as WebGL backend improvements for ONNX.js. Based in Sunnyvale, he has a strong systems background from earlier roles at Bing and Windows Store where he built scalable data pipelines and experimentation services. An active open-source contributor, Hariharan’s work touches core interoperability pieces that enable model portability and high-performance inference across platforms. He brings a blend of research experience from Purdue and practical infrastructure delivery at Microsoft, and has a knack for surfacing subtle correctness and shape-inference fixes that prevent downstream model conversion failures.
11 years of coding experience
7 years of employment as a software developer
Master's Degree, Computer Engineering, Master's Degree, Computer Engineering at Purdue UniversityBachelor of Engineering (B.E.), Electrical and Electronics Engineering, Bachelor of Engineering (B.E.), Electrical and Electronics Engineering at PSG College of Technology
English, Hindi, Tamil, German
Stackoverflow
106reputation
19kreached
6answers
0questions
Github Skills (24)
javascript10
c-language10
python10
operation10
tensorrt10
machine-learning10
onnx10
kernel10
deep-learning10
tensorflow10
kernel-mode10
cuda10
tensor10
cprogramming-language10
data-structure9
Programming languages (6)
Github contributions (5)
microsoft/onnxruntime
Feb 2019 - Jan 2023
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:2 releases, 1359 reviews, 1031 commits in 4 years
Contributions summary:Hariharan implemented support for the uint8 datatype for the Upsample operator, including integrating it into both CPU and CUDA execution providers. They also fixed a comment in the op_node_proto_helper, and integrated support for the valid `axis` attribute in the TopK operator implementation. Additionally, the user added a CPU kernel for the new Shrink operator. They also fixed a few small bugs across multiple areas and made other minor improvements to the core code base.
runtimetrainingtensorflowai-frameworkaccelerator
onnx/onnx
Mar 2019 - Jan 2020
Open standard for machine learning interoperability
Role in this project:ML Engineer
ML EngineerContributions:41 reviews, 19 commits, 65 PRs in 10 months
Contributions summary:Hariharan primarily focused on enhancing the `onnx/onnx` repository by modifying existing operators and adding new shape inference capabilities. Their contributions included implementing dynamic 'k' support for the TopK operator and fixing shape inference for the Slice operator. The user also added shape inference logic for the Expand and Tile operators, which is crucial for model conversion and execution. Furthermore, the user addressed various build breaks and performed code refactoring to improve the overall functionality of the repository.
pytorchmxnetdeep-learninginteroperabilitymachine-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.