Academic

The site you are reading: a single Next.js app that serves this website, the iPod Classic and iTunes from one SQLite database, with photos and videos kept on a Fly volume outside the repository. Every view reads the same data, so a new recipe or photo shows up everywhere at once.
Tech: TypeScript, Next.js, React, SQLite, Drizzle ORM, Docker, Fly.io, Vitest, Playwright.

A full reimagining of my personal website as a 1:1 iPod Classic, complete with a click wheel, keyboard support, mobile touch controls, and a tiny 320x240 screen for browsing around.
Features: Cover Flow browsing for guitars and photos, full-text articles, a 64-mug collection, YouTube videos, SoundCloud tracks, music recommendations, tweets, timeline and links, a Misc section, media that keeps playing while navigating, seeded content with live RSS and feed refreshes, and graceful fallbacks so the screen never starts blank.
Tech: TypeScript, Next.js, React, SQLite, Drizzle ORM, Zustand, Docker, Fly.io, Vitest, Playwright, and Web Audio for the click-wheel ticks.

A fun, weekend side-project jam: a single-page web app for visualizing what it costs to fill different vehicles, including motorcycles, across U.S. states and dates.
Features: interactive date scrubbing with an exact date picker, state-by-state lookup, support for multiple vehicle types and fuel types, transparent AAA/EIA source notes, "vs today" comparisons, and scheduled background refreshes with bundled seed data so the app still has something useful to show on a brand-new machine.
Tech: Python, Flask, SQLite, JavaScript, HTML/CSS, Docker, scheduled jobs, and upstream AAA/EIA data parsing.

As part of our advanced operating systems class, we designed and developed a distributed collaborative text editor, with the twist that multiple people cannot edit on the same line.
Features: designed and developed a custom "Chain" data structure based on our needs, and defined and implemented message semantics for the distributed environment. The edited file is saved to a local database, and crashes are handled.
Tech: Java Sockets, distributed systems concepts, Java Swing.

This movie-review website is a side project of me and four of my close friends, where the idea is that movies are reviewed with 5 POVs. The 5 authors can separately log in as admins and type their review in a simple form, which is then pushed to the SQL database. The end user simply browses through movies across multiple pages, reads their summaries, and then clicks on one of the 5 reviewers' reviews. Additional features include customized lists, search, and being optimized for both desktop and mobile.
This was entirely built from scratch by me, with design and business ideas from others. It was a really good way to use topics I learnt in college: development frameworks, software project management, database design, HTML, CSS, JavaScript and software architecture (MVC here).
It was a really enriching experience, where I learnt the value of proper communication and planning. It is on hold for the moment, but I am really happy with the end result.
Tech: Python, Django, Bootstrap, MySQL, PythonAnywhere (for hosting).

Artificial Intelligence (AI) has a huge scope in automating, streamlining, and increasing productivity of the music industry. Here, we look at AI-based techniques for classifying a piece of music into multiple genres and then converting it into another user-specified genre. Plenty of work has been done in classification, but using traditional machine learning models which are limited in accuracy and rely heavily on features to train the model. The novelty of this work lies in its attempt to convert the genre of music from one type to another.
This paper focuses on classification achieved by a model trained via Convolutional Neural Networks. Conversion of music genre, a relatively less worked upon field, is discussed along with implementation details. For conversion, we initially convert the input file to a spectrogram. A database of all genres is maintained at all times and a random file from the user-selected genre is also converted to a spectrogram. These spectrograms are processed and converted back to signals, and the user can listen to the converted audio file. Validation of the conversion was performed via a survey of end users. Thus, a novel idea of music genre conversion was put forth and validated with positive outcomes.