Working with Prof. Carlee Joe-Wong on performing Federated Learning on Graph Neural Networks for tasks such as Graph classification, Node classification, Node and Link prediction using personalized methods using PyTorch.
MathWorks
May 2021 - August 2021
Engineering Development Group Intern
Built a working prototype of MATLAB WebApps as a user authored custom dashboard on ThingSpeak.
Implemented an OpenID Connect Provider for user authentication using MathWorks account as a part of ThingSpeak to bridge the gap between the MATLAB WebAppServer and ThingSpeak.
HealthCloudAI
June 2019 - April 2020
Machine Learning Engineer
Developed a Dynamic Graph Convolutional Neural Network (DGCNN) model to predict clinical diagnosis from unstructured clinical text using Tensorflow to improve the quality and effectiveness of patient care.
Also developed a model to generate personalized questions and associated symptoms/risk factors based on the demographics and the medical history of patients.
Designed and implemented a pipeline consisting of acquiring the medical data, cleaning the data, training models, validating them and deploying them on the Google Cloud Platform.
University of Notre Dame
May 2018 - July 2018
Research Intern
Developed Generalized Identitification of Cohort Specific Themes (GIcST) framework to systematically extract themes differentiating texts of two generalized population sub-groups while accounting for overall population-level experiences.
This framework automates the process of discovery of psychological themes with respect to outcomes from unstructured psychological intervention texts paving the pathway for personalizing interventions and to gain insights into the practices surrounding patient conditions and outcomes, aimed to ultimately better inform the quality and effectiveness of care.
IIT Gandhinagar
May 2017 - July 2017
Summer Research Intern
Implemented Binary Image Segmentation in MATLAB by using the graph representation of Simple Linear Iterative Clustering (SLIC) superpixels of an image.
Analyzed different methods of Spectral Clustering and understood the graph representation of an image and compared this approach with the traditional K-means clustering.