Purvanshi Mehta
Co-Founder at Lica World
San Francisco, California, United States
Summary
👤
Senior🎓
Top SchoolPurvanshi Mehta is a machine learning researcher-practitioner and entrepreneur with a decade of experience building graph and multimodal learning systems, now co-founding Lica World while serving as a Founder Fellow at South Park Commons in San Francisco. She has applied her expertise across industry research roles at Microsoft and Amazon Lab126—working on graph intelligence, pre-training GNNs for NLP, and security-focused ML—and has a strong academic foundation from an MS in Computational Neuroscience and Statistical Machine Learning. Her open-source contributions include practical advances to Bayesian CNNs in PyTorch, reflecting hands-on work on probabilistic layers, inference tweaks, and data augmentation. She has a track record of translating research into impact, from integrating satellite-based wealth mapping into IPA’s Poverty Probability Index to publishing and presenting multimodal fusion and interpretability work. Known for combining probabilistic deep learning with graph techniques, she writes about her research and ideas publicly, making technical work accessible beyond academia.
10 years of coding experience
6 years of employment as a software developer
Master of Science - MS Computational Neuroscience | Statistical Machine Learning , Master of Science - MS Computational Neuroscience | Statistical Machine Learning at University of RochesterBachelor of Engineering Software engineering, Bachelor of Engineering Software engineering at Thapar UniversityB.E., B.E. at St. Mary's Con. Sr. Sec. School
Spanish, English
Stackoverflow
58reputation
6kreached
3answers
1question
Github Skills (15)
bayesian-network10
pytorch10
variational-inference10
convolutional-neural-networks10
data-augmentation10
python9
image-recognition9
pip6
stanford-nlp6
dependency-parsing6
nlp6
named-entity-recognition6
spacy6
stanford-parser6
java6
Programming languages (5)
Github contributions (5)
kumar-shridhar/PyTorch-BayesianCNN
Oct 2018 - Jul 2019
Bayesian Convolutional Neural Network with Variational Inference based on Bayes by Backprop in PyTorch.
Role in this project:
ML Engineer
ML EngineerContributions:5 commits in 9 months
Contributions summary:Purvanshi contributed to the Bayesian Convolutional Neural Network project by adding data augmentation techniques, including random flips and crops. Further contributions involved changes to Bayesian convolutional layers, specifically the `BBBConv2d` and `BBBLinearFactorial` layers, indicating work on the model's architecture and inference mechanisms. They also resolved conflicts and made modifications to the model's beta, as well as overall adjustments to the environment. This suggests a focus on implementing and refining the Bayesian CNN model within the PyTorch framework.
bayesian-inferencebayesian-networkspymc3variational-inferencetemporal-data
purvanshi/isolvemath
Mar 2017 - Feb 2018
Contributions:1 release, 34 commits, 4 PRs in 11 months
englishproblemsmathsenglish-language
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.