Niharika D'souza is a Research Scientist at IBM Research AI with a decade of experience developing geometric deep learning and multimodal fusion methods that bridge graph signal processing, information theory, and statistical learning. She has led work on multidimensional and multiplexed graph neural networks—recognized at MICCAI and published in Medical Image Analysis—and pioneered information-theoretic formulations that enable end-to-end optimization of multi-graph structured representations. Her recent contributions include physics-aware GDL for radial imaging that delivers clinical-grade measurements in real time, plus practical systems like phrasally grounded fact-checkers for radiology reports and LLM-based prompt optimization for structured discovery. Trained with a PhD from Johns Hopkins and a strong mathematical background, she combines rigorous theory with translational impact, holding multiple patents and IBM awards for advancing AI for multimodal fusion. Notably, her work emphasizes resource-efficient models that make high-precision clinical tools feasible outside large compute environments.
10 years of coding experience
5 years of employment as a software developer
Johns Hopkins University
Bachelor of Technology (B.Tech.), Electrical Engineering, 9.17/10.00, Bachelor of Technology (B.Tech.), Electrical Engineering, 9.17/10.00 at IIT Kharagpur
Contributions:58 commits, 7 PRs, 1 comment in 25 days
stanfordmatlabautoencodersstanford-onlinesparse
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.