Aditya Rathod

I’m an undergraduate CS student who loves
 

Hello, my name is Aditya Rathod and I'm an undergraduate at The University of Texas at Dallas. ☄️

My technical interests include data science, machine learning, inferential statistics, and deep learning. Most of my projects revolve around these core interests, but I've also done some React/Vue webdev and Oculus VR app development before.


Projects

Liform

Medical + data + analytics app.

Details coming soon.

Reportik

Abstractive news article summarization using sequence-to-sequence models.

Used Python, TensorFlow, Keras, BeautifulSoup, FastText, and other tools to collect a dataset of ~26k CNBC articles, along with their summaries. Repurposed LSTM-based neural sequence-to-sequence language model as described in Sutskever et al. to the domain of long-form text summarization.
Findings: Seq2Seq models cannot formulate coherent summaries of long spans of text.


Project code →

AR Sandtable

Augmented reality sandbox with topography simulation.

AR Sandtable is an implementation of a project developed by a lab at UC Davis. Working on this project helped me learn about the C++ compilation process, Makefiles, modifying GLSL shader code, and shell scripting. This was a team project, and we extended the codebase by adding tides and one-click water to lava shader switching.


Photo and discussion →

FLEX

Progressive Web App to display appointments with teachers.

FLEX Time is a period in which students can book appointments with their teachers for additional help, test make-ups, or other activities. Appointments are booked electronically through a website which is hard to use and not designed for mobile devices. My app provided a one-touch, easy-to-use interface to book and display appointments. Other features included one-tap check-in to students' appointments and automatic login for more convenient use.

Technologies used include Vue.js, NodeJS, and Express.


Client Repo / Server RepoProject Description →

The Forge

Parts inventory system for Irvington Robotics Academy.

Worked on building a parts inventory and check-out system for Irvington Robotics Academy, of which I was an officer. The final web application will use React and Redux on the client side and Firebase on the backend. The goal was to build a integrated system for running our club, starting with the tedious task of inventory management. Losing or loaning parts and never getting them back puts a strain on our student-run Academy, so having a way to see where each part has gone will increase club efficiency and reduce unnecessary expenditures.


Note: the link below is to our alpha deployment, written in Mithril.js. The React/Redux branch is still under development.


Alpha Deployment →

VRCade

First-person arcade game for Oculus Rift.

As part of a Principles of Engineering final project, I created a virtual reality version of Pac-Man to learn VR and game development. A friend created the in-game art assets, while I worked on all other aspects ofthe project. The game is a fast-paced, first-person arcade-style experience and was exhibited at Maker Faire Bay Area 2018. It was also featured in the October 2018 issue of Make Magazine.


Development Log →

HQ Trivia QAS

Proof of concept of an OCR-based pipeline to answer HQ Trivia questions.

HQ Trivia is a popular smartphone quiz game played by thousands of people worldwide, in which players compete for a cash prize. However, since the game runs on a smartphone, there is a chance for exploitation by bad actors. I designed a proof-of-concept system to capture, transcribe, and search up answers to HQ Trivia questions in realtime. I then detail methods to thwart cheaters by designing OCR-proof interfaces. Ethics Note: All testing was performed on captured video gameplay, and NOT on live matches.


As HQ Trivia gained popularity, my article brought up valid concerns about the viability of a high-stakes quiz game delivered through a mobile medium. Outlets such as The Verge, The Washington Post, and TechCrunch wrote about and linked to my suggestions for the app to add OCR-proof interface elements and detailed my proof-of-concept.


Blog Post →

Contact

javascript needed to reveal email

LinkedIn Profile