Tanushree Burman

Hi! I am a second year Computer Science Ph.D. student at Tufts University. I am a member of the Multimodal Learning, Interaction, and Perception Lab, led by Jivko Sinapov and I work as a Research Assistant at the Tufts Center for Engineering Education and Outreach. My interests include reinforcement learning, robotics and AI/ML in STEM education. I am passionate about developing technology that will introduce AI concepts to students and demystify AI models through a hands-on learning approach using robotics.

During my undergraduate education at Yale University, I worked with Kim Blenman in the Blenman Innovation Group and with Julian Jara-Ettinger in the Computational Social Cognition Lab. I have also been fortunate enough to intern as an AI/ML engineer in the Lockheed Martin AI Center.

Email  /  Github

profile photo
Projects
blind-date Smart Motor and Smart App System

In this project, I developed a web-based User Interface that we call the Smart App. Children can use the Smart App to learn supervised machine learning techniques using a low-cost hardware component. We call this Lego-compatible component a Smart Motor which contains a servo motor, a sensor port and a microprocessor to allow for programming different training algorithms. This UI allows students to train and test the Smart Motor, add and delete individual training data points and view more detailed visualizations of the AI model. When students let the Smart Motor run on the training values they inputted, the graph highlights the nearest data point that is influencing the decision making of the algorithm. Through these visualizations, we hope children are able to have a more intuitive understanding of how the algorithm is making underlying decisions.

Publications
blind-date Prediction of Death after Terminal Extubation, the Machine Learning Way
Jessie Huang, Dennis Shung, Tanushree Burman, Smita Krishnaswamy, Ramesh Batra
Journal of the American College of Surgeons, 2022
Paper

In this paper, we use an LSTM to accurately predict death within 2 hours post-Terminal Extubation and compare to the standard regression model on priori list of clinical variables. We found that our LSTM was superior to clinical predication score. This information will improve organ donation rates and in facilitating End of Life discussions.


Website inspired by Jon Barron.