Projects
Research Experience at The University of British Columbia, Canada
My research experience as a visiting student at The University of British Columbia, Canada would be one of my most memorable experiences. I completed my undergraduate thesis project in ML under the guidance of Prof Apurva Narayan. The project was highly creative, intellectually challenging, and aimed at futuristic practical applications in the domains of Human Intent Recognition (HIR) and Human Activity Recognition (HAR) using time-series analysis. I created a novel synthetically crafted pseudo-realistic dataset for HIR from an existing HAR dataset and proposed a novel light-weight deep learning architecture of dilated CNN(dCNN).
Note: Paper accepted at IEEE,SMC 2020 Toronto.
Internship at Vedalabs
I developed a currency recognition system on an embedded system that could accurately predict currencies within a few seconds. I worked on different image processing techniques to improve the effectiveness of the currency dataset and used the object detection algorithm You-Only-Look-Once (YOLO) for the deep learning model. The final system was able to achieve an accuracy of 99.54%.
This internship helped me in gaining technical skills but also gave me an opportunity to experience the creative work culture of a thriving startup. It also gave me an insight into how AI could be used for practical applications.
*Upcoming Project
Reseach Project at Prof Shie Mannor's Reinforcement Learning Research Lab, Technion, Israel
The project focuses on the topic of Safe Reinforcement Learning and aims to bring a theoretically grounded paradigm KWIK (knows what it knows) into practice. We aim to develop a generic framework for different environments that would make training agents safer by incorporating the principles of model uncertainty and confidence bounds to determine safe actions.
