Moving Target Interception using Multi-Agent Reinforcement Learning (MARL)

May 2023

• Developed and published a MARL framework, training agents to make co-ordinated decisions to capture an evasive thief.

(Python, Numpy, OpenCV)

Music Recommendation System

April 2023

• Achieved the highest Mean Average Precision (mAP) of 0.78 among NYU projects by developing and hyperparameter tuning a music recommendation system on interactions of user and song.

(Spark, Dask, Python, Hadoop)

Analysing Optimal Video Game Playing Conditions

January 2023

Performed predictive and descriptive analysis on Video Game FPS dataset and wrote a report on it (under Dr Pascal W).

• Collaborated in a team of 4 to perform an inference test (Kolmogorov–Smirnov Test) to check if Moore’s Law is still valid.
• Trained a neural network to predict FPS (RMSE 0.1025), built a model to recommend optimal settings to play a video game on.

(TensorFlow, Scikit-Learn, Scipy, statsmodels, LightGBM)

Using Retinal Fundus Images for Multi-Disease Detection

October 2021

Performing ensemble Deep Learning on Retinal Fundus Images for Multi Disease Detection (under Prof. Sarada S).

• Collected over 6000 retinal fundus images from an eye hospital in Mumbai.
• Trained an ensemble CNN to identify 45 diseases from collected images with an accuracy score of 0.97

(TensorFlow, CNN, OpenCV)

swachhdata

August 2020 - Present

Swachhdata is a python library that provides tools for data, text, and image preprocessing with over 50,000 downloads worldwide.

• Conceptualized swachhdata and coded 20+ functions for data, text and image preprocessing.

(Regex, Git, PyPi, NLTK, OpenCV, Gensim, NumPy, and Pandas)

Varta Twitter Recommendation

August 2021

A Twitter website where any user can log in to the website using Google Login to receive personalised Twitter.

• Programmed a model which used Pearson Correlation Score and Collaborative Filtering to give Twitter recommendations.

(Scikit-Learn, Git, NLTK, NumPy, Pandas, and Flask)

Wherebnb

October 2020

A website that predicts the price and popularity of a listing using ML and provides an interactive map of real listings.

• Trained machine learning models with an accuracy score of 0.932, and programmed the backend in Python using Flask.

(Python, Flask, TensorFlow, Scikit-Learn, HTML-CSS, and Tableau)

NLP Cleaning Pipeline

July 2020

A website that lets a user test and learn about different NLP cleaning procedures.

• Wrote functions for different cleaning procedures, and programmed the backend in Python using Flask.

(Python, Flask, NLTK, Emoji, TextBlob, HTML-CSS)