Rohan Mukherjee
Applied Scientist at Amazon
California, United States
Summary
👤
Senior🎓
Top SchoolRohan Mukherjee is an applied scientist with nine years of experience driving ML model deployment and compiler-level performance engineering, currently at Amazon in California. He holds a Ph.D. and MS in Computer Science from Rice University, where his thesis explored the role of context in program search and synthesis, blending theoretical rigor with practical systems work. At AWS and Amazon he has focused on taking deep learning models to custom hardware and production, and his open-source contributions to the widely used TVM compiler include TensorRT integration and dynamic-shape optimizations for object detection—work that improved serialization, testing, and runtime performance. His background in electronics and game-theoretic modeling for energy trading underscores a multidisciplinary approach to problem solving, pairing hardware-aware ML engineering with a strong research foundation.
9 years of coding experience
Nava Nalanda High SchoolMaster of Science - MS, Computer Science, 3.871/4.00, Master of Science - MS, Computer Science, 3.871/4.00 at Rice UniversityBachelor of Engineering (B.E.), Electrical, Electronics and Communications Engineering, First Class Honours (CGPA-9.07/10.00), Bachelor of Engineering (B.E.), Electrical, Electronics and Communications Engineering, First Class Honours (CGPA-9.07/10.00) at Jadavpur University
Hindi, Bengali, English
Github Skills (13)
tensorrt10
compiler10
relay-modern10
tvm10
machine-learning10
deeplearning-ai10
compiler-compiler10
performance-monitor10
performance-analysis10
deep-learning10
gpu10
python10
relay10
Programming languages (3)
Github contributions (5)
apache/tvm
Oct 2020 - Jul 2021
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
ML Engineer & Performance Engineer
ML Engineer & Performance EngineerContributions:18 reviews, 7 commits, 14 PRs in 9 months
Contributions summary:Rohan contributed to the TVM compiler stack by implementing support for TupleType inputs in the CheckReshapeOnly pass within the Relay virtual machine. They also worked on integrating and optimizing TensorRT within the TVM framework, focusing on dynamic shapes for object detection models and addressing serialization issues. Their contributions involved refactoring code, adding tests, and debugging to improve the performance and compatibility of the TVM with TensorRT.
metalvulkancompilertensoropencl
rohanmukh/ELEC677
Sep 2016 - Jul 2021
Contributions:4 pushes, 2 branches in 4 years 10 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.