Ishana Shastri
Engineer at Doppel
Cambridge, Massachusetts, United States
Summary
🤩
Rockstar🎓
Top SchoolIshana Shastri is an engineer and MIT master's student in Computer Science and Applied Mathematics with nine years of experience building ML systems and research-driven software. She has applied her skills across industry and academia—from MosaicML (contributing to the widely used Composer training library) and Alphabet's Wing to CBMM and IBM Research—focusing on model training efficiency, instrumentation, and applied computer vision. Her work spans production-oriented ML engineering and hands-on research, including medical imaging predictors and tooling improvements like in-memory logging, GPU device selection, and decoupled save/load logic. Based in Cambridge, she combines deep technical rigor with practical product sensibilities and a side interest in AI alignment research, reflecting a blend of long-term safety thinking and shipping-oriented engineering.
8 years of coding experience
2 years of employment as a software developer
Master's degree Computer Science and Artificial Intelligence, Master's degree Computer Science and Artificial Intelligence at Massachusetts Institute of TechnologyPoolesville High School
English, Spanish
Github Skills (9)
pytorch10
machine-learning10
deeplearning-ai10
deep-learning10
trainings10
python10
modeling10
mlops9
nlp8
Programming languages (2)
Github contributions (5)
mosaicml/composer
Jun 2022 - Aug 2022
Supercharge Your Model Training
Role in this project:
ML Engineer
ML EngineerContributions:72 reviews, 8 commits, 14 PRs in 2 months
Contributions summary:Ishana contributed significantly to the development and improvement of the `composer` library, a framework for machine learning model training. Their work included integrating an `InMemoryLogger` for tracking training data and applying label smoothing for more consistent accuracy results. Furthermore, the user made modifications enabling GPU device selection and decoupled saving and loading logic within the GLUE entrypoint, suggesting a focus on improving model training efficiency and flexibility. They also updated the metrics handling.
pytorchml-systemsdeep-learningneural-networksmachine-learning
ishanashastri/composer
Jun 2022 - Aug 2022
library of algorithms to speed up neural network training
Contributions:110 pushes, 10 branches in 2 months
speeddeep-learningneural-networksmachine-learningtraining
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.