Competitions
 | Data Science Track We bagged the 2nd place out of more than 3500+ teams for the task of predicting the time to default for a credit card customer. Utilising a gradient boosting ensemble, we were able to achieve an accuracy of 82.4%, battling the heavy dataset imbalance present. |
 | Published and presented at ICCR 2022 Conference Proceedings [paper] [code] We achieved 1st place internationally for the task of real-time navigation and manipulation. Our novel contributions were designing a probabilistic travel distance minimization algorithm for multiple waypoints traversal in the shortest time and developing an adaptive Model Predictive Control strategy. |
 | We bagged the bronze medal among 23 competing IITs in the high prep event. The aim was to develop a website to help train and monitor a model on a traffic sign recognition dataset, with the capabilities to handle the dataset imbalance and understand model inference. Utilising the MicronNet architecture, we were able to achieve an accuracy of more than 98.7% on the GTSRB dataset. |