This is a Java Swing based desktop application that allows users to create and manage conferences. The application worked like a LinkedIn for conferences and users are able to browse conferences, look at various events within the conference, connect with other people in the same events and chat with connections.

The application was built using SOLID principles and various design patterns like dependency inverstion, factory design pattern etc.

The project was completed in a team of 7 developers as the final project for the software design course(CSC207) at UofT.

A research project where we modified the architecture of Speech Enhancement GANs (SEGANs) and evaluated their performance in the task of sound enhancement on different datasets.

We experimented with the Wasserstein loss metric and the 'Departure from Normality' metric (a Schur decomposition based similarity metric for LSGANs) and found some interesting results which were summarised in a paper.

The project was completed in a team of 3 students as the final project for the deep learning course(CSC413) at UofT.

A common issue I had with outlook was that I would miss events because the sender did not include a calender invite to the invite and just accepting the event did not automatically add it to my outlook calender. I got tired of it and tried to look for solutions in an extension but I could not find an extension that could scan emails for events and prompt me to add them separately.

I found out that Microsoft has outlook dev tools that allow devs to create their own extensions so, along with my friend, I decided to build my own extension. We used Open AI's GPT-3.5 to parse emails and identify events in outlook emails and the outlook dev tools to add the events to the user's calender. This was a fun project that we use ourselves to improve our productivity.

This is a JavaScript library for visualisations related to the game of football(or soccer) and fantasy football. It allows users to visualise players on the pitch, make gameplay animations, compare team strengths in FIFA and compare different players' in game attributes in FIFA.

The library was developed using vanilla JS and was deployed with documentation on Heroku. The backend for the host website was build using NodeJS.

The project was completed in as the final project for the web development course(CSC309) at UofT.

This is a tinder like web app that allows users to find people to play sports with and connect with them. It also allows users to create and manage sports tournaments. The app was built using HTML and CSS for the frontend, NodeJS for the backend and Mongoose for the database.

The project was completed in a team of 5 developers as the final project for the web development course(CSC309) at UofT.

I was interested in Neural Style Transfer and thought that it would be interesting to see if we can extend it to audio samples as well. So I proposed to lead a UTMIST (University of Toronto Machine Intelligence Student Team) project to explore the same.

We followed a paper that used CNNs on the gram matrices of the spectrograms of audio samples to convert the "style" of a content audio into the the style of a target style audio. The results were hard to interpret but we learned a lot both about audio files and CNNs from this project and presented our results in the year-end showcase.

The project was completed in a team of 5 developers as the UTMIST ML project for 2021.

After constantly getting my food cold before deciding to watch something on netflix on a weekend, I was tired and decided to create this app.

This is a webapp that uses GPT-3.5 to provide movie recommendations based on prompts. So you can ask it to recommend you "a critically acclaimed horror movie that I can watch in 1 and a half hours" and it will give you three recommendations with their names, release date, imdB score and a short summary to help you decide between these movies.

I used OpenAI API for the "recommendation engine" and JS and NodeJS for the webapp. An interesting part of this project was just figuring out how to extract the movie data from imdB using cheerio ( because imdB no longer allows access to their APIs :( )

I re-created the classic doodlejump game in MIPS Assembly. This project, while sometimes frustrating, was a very good learning experience and gave me a lot of respect for low level languages.

Moods is a Journal app that uses sentiment analysis to detect mood levels of the user and makes it easier for them to keep track of their mental health. The app also displays useful statistics that help users identify what factors affected their mental health the most over time (positively and negatively). The front end was implemented using flutter and the backend was implemented using Python Flask.

This project was comleted in a team of 4 developers for UofT Hacks 2021.