Biography
Pramit graduated from the University of British Columbia, Vancouver, Canada, with an MASc degree in Electrical and Computer Engineering in February 2021. As a MITACS Globalink Research Fellow and Graduate Research Assistant in HCT Lab, his work was primarily targeted towards finding different alternative vocal communication pathways for people with speaking disabilities, by employing deep learning strategies to map our active thoughts or gestures to the acoustics space for artificial voice synthesis.
Supervised by Prof. Sidney Fels, Pramit's master's thesis focused on investigating information-theoretic view of speech-related motor control. Using deep learning-based strategies, he particularly addressed the question of whether humans can take advantage of categorical speech perception as a constraint to reduce the difficulty of a hand-to-speech task. During his master's, he also worked on developing gesture-based speech interfaces and EEG-based imagined speech recognition systems. He was awarded the prestigious Young Scientist Award by MICCAI Society in 2020.
After completing his master’s, Pramit started studying for a DPhil under the supervision of Prof. Alison Noble. His doctoral research aims at solving the challenges encountered in the federated learning scenario like class imbalance, non-IID data distribution, etc., thereby facilitating multi-institutional collaborations without sharing patient data in the medical image analysis domain.
Research Interests
- Deep Learning
- Medical Image Analysis
- Computer Vision
Current Projects
Pramit is currently working on the CALOPUS project which is a collaborative research project between the Department of Engineering Science and Women’s Reproductive Health at the University of Oxford and the Translational Health Science and Technology Institute (THSTI).
The project brings together an international team of experts in clinical and engineering science to develop a point-of-care ultrasound solution that will identify whether pregnant women are at risk of an adverse pregnancy event. The cross-country data sharing restrictions act as an impediment in developing a common model effective in both centers. Pramit's research investigates the ways to enable model training without requiring any data sharing among the centers by employing federated learning algorithms.
