PEFT Strikes Back - Exploring Efficient Finetuning of Language Models
Integrated verbalizers with LLMs and compared multiple PEFT techniques for efficient alignment of LLMs for text classification tasks.
Apr 30, 2024
A Multi-Stage Vision-Language Framework for Knowledge-based Visual Question Answering
Devised a two-stage VLM-based pipeline by utilizing knowledge of LMs to first sample multiple candidate answers and then using CLIP to select the most likely choice.
Dec 5, 2023
Predicting FIFA World Cup Outcomes
Predicted FIFA World Cup groups and matches using 10 different machine learning algorithms including ensemble methods.
Nov 30, 2023
Drowsiness Detection in Drivers
Designed a CNN-based neural network to detect drowsiness in drivers frame-by-frame using RGB videos.
Jul 5, 2020
Design of Model Predictive RBFN Controller for Non-linear Plants
Utilized Simulink and MATLAB to design a RBFN controller for non-linear plants modelled using a U-model.
Jun 28, 2020
Private Chat Application using MongoDB and Socket.io
Implemented a multi-client chat system along with messages retention feature.
Jun 20, 2020
News-text Classification using a Weighted RNN
Implemented weighted RNNs, providing an alternative to attention mechanism by dynamically weighing each token based on their impact on determining the class of the news article.
Nov 28, 2019
Leakage Detection in Smart Water-Distribution Systems
Experimented with test-bed to collect experimental leak data using various sensors and utilized it to train a neural network that can predict transient leaks in the system.
Nov 28, 2019