Project part of Research Internship at Natural Language Processing Lab, Virginia Tech.
Duration: May 2021 - Present
Mentor: Prof. Lifu Huang, Dept. of Computer Science, Virginia Tech
Details
Preparing a benchmark dataset for Natural Language Understanding tasks which can be derived based on the WikiHow resources. Designing Neural Baseline Models for this task using BERT, VisualBERT Baseline 1: Modelled as binary classification problem for each set of task description and entities as input using pre-trained model BERT.
This project is a hackathon submission to Uber’s Hackathon Hacktag 2021. The hackathon required providing an innovative solution around the theme “Uber beyond Mobility and Delivery."
Duration: February 2021 - March 2021
Ideation Details: Urban Consolidation Centers(UCCs) are locations, typically on the outskirts of cities, where deliveries are brought, sorted, and then dispatched. Goods from multiple suppliers can be consolidated into fewer shipments, making it possible to optimize loads and truck sizes, thus cutting down on the number of trips and vehicles required.
Developed a Novel Learning Framework Broad Class Incremental Learning System to tackle the forgetting problem of Class Incremental Learning
Project part of Summer Internship 2020 at IIM Ahmedabad.
Duration: April 2020 - July 2020
Mentor: Prof. Anuj Kapoor, IIM Ahemdabad
Details:
Compared the service quality ratings of hosts who own multiple properties to those who own only a single property, using detailed property-level data from Airbnb from the cities of Amsterdam & Antwerp Employed a Difference-in-Differences Design to causally study the effect of government crackdown on hosts who own multiple properties and disentangle the effect on service quality rating of hosts Used Latent Dirichlet Allocation (LDA) to classify 15k guest reviews into Property based & Host based and hence established a shift from Property to Host based reviews after the crackdown Investigated heterogeneity in effects using data driven machine learning algorithm: Causal Forest Paper Revised and Re-submitted to Elsevier’s IJRM.
Developed a new method for tumor segmentation from brain MRI using computationally less intensive clustering algorithms
Introduction:
The project was part of the Science and Technology Council’s Summer Project at IIT Kanpur. This project’s goal was to create an interactive game that could be played with gestures and to create a Game AI. This project was implemented in three stages.
Duration: May 2019 - July 2019
Details:
Stage I: Developed a space shooter game using PyGame and built a GameAI using Behaviour Tree Stage II: Incorporated gesture control for the game using OpenCV and Depth Sensing using Intel RealSense Stage III: Built a Game AI using Q-Network for the PyGame based space shooter game.