Jan '23 - Jul '23

MARBLER: Multi-Agent RL Benchmark and Learning Environment for the Robotarium
An Open Platform for Standarized Evaluation of Multi-Robot Reinforcement Learning Algorithms.
Code Demo Blog Paper

Jan '23 - May '23

Document Filtering for Drug Repurposing and Clinical Meta-Analysis
Guide: Prof. Cassie Mitchell, Pathology Dynamics Lab, Georgia Tech
Engaged in creating the BioSift dataset for drug repurposing and developed a set of NLI-based Zero-shot Multi-label Classifiers for a document classification benchmark.
Paper

Aug '22 - Nov '22

Biomedical Entity Linking Survey
Guide: Prof. Cassie Mitchell, Pathology Dynamics Lab, Georgia Tech
Conducted a comprehensive survey of Biomedical Enity Linking (BioEL) models and created a benchmark of uniformly formatted datasets.
Paper

Aug '21 - Jul '22

Joint Completion and Alignment of Multilingual Knowledge Graphs (AlignKGC)
Guide: Prof. Soumen Chakrabarti, Department of Computer Science, IIT Bombay
In the domain of multilingual KGs, Knowledge Graph Completion (KGC) or fact prediction adds valuable information for the Relation Alignment (RA) and Entity Alignment (EA) tasks, and vice versa. The AlignKGC framework aims to jointly learn KGC, EA and RA for efficient knowledge transfer across KGs.
Report Paper

Jan '21 - Apr '21

Entity Alignment in Multilingual Knowledge Graphs
Guide: Prof. Soumen Chakrabarti, Department of Computer Science, IIT Bombay
Entity Alignment (EA) plays a vital role in automatically integrating multiple knowledge bases. Has BERT-INT, which gives almost saturation results on DBP15K, solved the Entity Alignment problem?
This work demonstrates the inadequacy of BERT-INT for EA in the presence of low-resource pairs or high entity multiplicity.

Mar '21 - Apr '21

Online Reinforcement Learning for Lane Following
Guide: Prof. Shivaram Kalyanakrishnan, Department of Computer Science, IIT Bombay
Lane following is a crucial part of the Autonomous driving problem. Is it possible to learn lane following in an online and efficient fashion using a very low-dimensional discretized state-space?
Report

Fall '20

Automatic Raga Recognition in Indian Classical Music
Leveraged tonic normalized pitch-tracked frequencies of music samples for raga classification using LSTMs with attention.
Code Report

Fall '21

Groot - The Urban Vertical Farming Unit
Designed a smart Modular Vertical Farming unit for climate-independent agriculture in compact spaces. Also designed a Farm-as-a-Service platform in the form of an android application. Conducted detailed mechanical and business analyses.
Demo Report

Fall '21

Profile Recommendation System for Online Dating
A machine learning algorithm to rank profile-pairs by generating compatibility vectors for them. Uses proxy-labeling to generate supervision signals for model training.
Code Report

Fall '21

Web corpus indexing and compression
Developed an inverted index for a corpus of 50,000+ web documents by encoding D-gaps. Compared various index compression techniques like Elias Gamma Coding, Golomb Coding, and Arithmetic Coding.
Code

Fall '20

Image Quilting for Texture Synthesis and Transfer
Implementation of a patch-based algorithm to synthesize texture from a given sample. This quilting algorithm is also extended to transfer a given texture to any target image.
Code Report

Fall '20

Shortest Path in a Maze
Shortest path algorithm for maze solving that models a 2D maze as a Markov Decision Process (MDP) and generates optimal policy using value iteration.
Code

Spring '21

Mastering Atari Games using Deep Reinforcement Learning
Deep Reinforcement Learning agent capable of surpassing human performance on classic Atari games like Pong, Breakout and Boxing using high-level sensory information in the form of game screen pixels.
Code Report

Summer '20

Star Shooter! (Video Game)
A Star Wars themed shooting game developed using PyGame to render graphics, animation, and sound.
Code