Jasha Droppo
Senior Principal Scientist at Amazon
Greater Seattle Area United States
Summary
🤩
Rockstar🎓
Top SchoolJasha Droppo is a Senior Principal Scientist based in the Greater Seattle Area with 11 years of industry experience and a Ph.D. in Electrical Engineering from the University of Washington. He brings deep research-to-production expertise from a long tenure at Microsoft followed by leadership in applied science at Amazon, specializing in speech and language and scalable deep learning systems. His open-source contributions include performance and algorithmic optimizations to the widely used Microsoft Cognitive Toolkit (CNTK), adding features like logsoftmax and RMSProp support for dense matrices. Known for blending rigorous academic training with pragmatic engineering, he often uncovers non-obvious performance wins in core numerical kernels that translate into real-world throughput improvements.
11 years of coding experience
19 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Electrical Engineering, Doctor of Philosophy (Ph.D.), Electrical Engineering at University of WashingtonBSEE, Electrical Engineering, Honors Program, BSEE, Electrical Engineering, Honors Program at Gonzaga University
Github Skills (15)
algorithm10
cuda10
numerical-optimization10
code-optimization10
algorithms10
machine-learning10
rnn-model10
c-language10
deep-learning10
cntk10
cprogramming-language10
n10
optimisation10
optimization10
distribute9
Programming languages (1)
Github contributions (2)
microsoft/CNTK
Aug 2014 - Oct 2017
Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit
Role in this project:
Back-end Developer & Algorithm Optimization Engineer
Back-end Developer & Algorithm Optimization EngineerContributions:196 commits, 50 pushes, 12 branches in 3 years 2 months
Contributions summary:Jasha's contributions focused on optimizing the CNTK (Microsoft Cognitive Toolkit) deep learning framework. They primarily modified code related to the softmax functions within the matrix classes, implementing logsoftmax and identifying performance improvements. The user also introduced new functionality for RMSProp training algorithm for dense matrices, and added functionality for a Log Computational Node. They also made changes to address build issues and incorporate new features.
pytorchpythondeep-learningc-plus-plusmachine-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.