Welcome!

About me

I’m a graduate student at Northeastern University, in the MS Data Science program.

Currently, I work at Fidelity Investments as an Associate Data Scientist. My work consits of leveraging unstructured text data to produce actionable insights which are used to drive business decisions.

My undergraduate studies were in Mumbai University, majoring in Information Technology. A Data Science enthusiast, I worked as a Data Science Intern at Hashtag Loyalty for a year. My work consisted of developing various Machine Learning models to predict and understand user behaviour, delivering presentations, and interating with the Business Development team to get clarity on client requirements.

My areas of interest include Natural Language Processing (working with problems like word sense disambiguation, pronoun resolution, summarization, etc.), and understanding and solving real life Data Science problems to provide monetary value to businesses.

I started solving competitions on Kaggle in my second year of undergraduate study. Starting from the very basic Titanic Survival prediction challenge to completing the Toxic Comments challenge with a bronze medal, it’s been a long way.

Details can be found in my Resume

Research:

• Deep Learning Based approaches for Recommendation Systems Link to paper

Recommendation systems are the most widely used Machine Learning algorithm in the industry. Recently Deep Learning algorithms have been successfully applied in fields like Computer Vision, Natural Language Processing etc. and have started being applied to Recommendation System. In this research paper we have studied different Deep Learning methods for Recommendation Systems.

• Facial Emotion Recognition using Deep Learning Link to paper

Human beings rely a lot on non-verbal communication and facial emotion is a large part of it. In this review paper we have covered the datasets and algorithms that are used for Facial Emotion Recognition (FER). The algorithms range from simple Support Vector Machines (SVM) to complex Convolutional Neural Network (CNN). We explain these algorithms through the fundamental research papers and go through their application to the task of FER.

Projects:

Here are some of the projects I’ve worked upon during my undergrad.

• Predicting Stock Movement Using News Headlines Code

The aim was to predict the movement of the stock price using the news headlines and past DJIA index data. Through the course of the project, various techniques have been explored, analyzed and formulated them to build the best model for our case. This project was done in the DS 5220 (Supervised Machine Learning) Class of Northeastern University, Spring 2020.

• Item Classification Code

I have classified a database of food items into appropriate segments (veg, non-veg, drinks, etc) using an unsupervised approach, since there are no features in our data, only (item_id, item_name).

I have used Word2Vec, SpaCy, GloVe and cosine similarity matrices.

• Web application for database operations Code

Extension of my project at Hexaware Technologies Ltd. It can perform CRUD (Create, Read, Update, Delete) operations on any database on a web app. Added a module to import a external database as a csv file which directly gets added/updated to the main database after validation. Also added a new custom self-designed CSS3 theme.

• Toxic comments classification challenge (Kaggle) Code

Built a model capable of detecting different types of toxicity like threats, obscenity, insults, and identity-based hate. Trained an ensemble model consisting of algorithms like LSTM, Attention Model, Glove embedding using Keras to get an accuracy score of 0.9865 on the leaderboard.

• Movie Recommendation System Code

Built an ensemble model for a personalized movie recommendation system. The ensemble model has content based, popularity based and collaborative filtering recommendation algorithms. Used the MovieLens datasets.

Contact

You can reach out to me on:

Email: diwan.p@northeastern.edu

LinkedIn: Profile

Phone: +1 (857)-472-1806