Recent content on Aditya Rathod
Built a local web dashboard and CLI for ASUS routers that shows all devices on the network and lets you block or unblock their internet access with one click. Used Claude to reverse-engineer the router’s undocumented API.
Started to rank every Chipotle location nationwide by price and user reviews, ended up building a Rust client for retrieving menus and locations for Chipotle’s reverse-engineered API instead.
Trained a bot on my message history to speak like me in a weekend, with blog post explaining how I did it.
Creating a chaotic personal AI bot by fine-tuning a language model on Discord messages.
A tiny library to implement Snowflake IDs in Golang. Mainly a means for me to practice idomatic Golang :)
Developed information retrieval pipeline for a League of Legends lore Discord chatbot. Used TF-IDF and SVD to generate a dense vector space model of the corpus for document retrieval. Leveraged DistilBERT-based model for extractive Q&A and FLAN-T5 finetuned model to generate conversational answers to queries on the corpus.
Worked in a team to build and pitch a solution in 24 hours that uses React and the OpenAI API to generate detailed travel itineraries based on customers’ preferences, even enabling them to book flights based on the itinerary. First place winner of the American Airlines challenge at TAMUhack 2023.
Developed a Python chatbot with a React frontend that can answer open-ended knowledge-based questions about data structures & algorithms and class administrivia.
Collaborated as part of the HackUTD Experience team to develop and present an introductory backend workshop to 50+ students. We built a Wordle clone with a leaderboard using Firebase :)
Worked in a team to build and pitch a solution in 36 hours that uses React Native and a machine learning model to track eye movements and display trends of fatigue during critical situations. First place winner of the Lockheed Martin and Children’s Health challenges at HackDFW.
Developed data pipeline and analysis tools for an AWS networking product. Dove deep, learned how networks worked from scratch, and built some cool tools that impact customers. That’s about as much as I can say. I now work full-time on the same team!
React + HTML canvas based robot simulator that allows users to create and control simulated robots in a 2D grid. Developed motion planning logic and created an interpreted robotic simulation language.
Joke API to generate repetitive strings of text (i.e. “jaja”). Primarily used to boost team morale in League of Legends games.
Basic React-based dynamic, animated broadcast graphics toolkit. Used in production for self-directed live in-person and streamed events, of sizes ranging from 500 to 1k+ total attendance. See it in action (00:00 to about 05:30): Another demo: Here’s the stream dashboard with scene groups (set of scenes played in sequence on repeat) and now-playing indicator toggles (that change content on…
Developed large-scale log event ingestion and query service leveraging C# and Elasticsearch. The goal was to make a more flexible, hackable alternative to Splunk. As part of the project, I also developed a novel query language and query parser to enable users to query logs using simplified syntax. I spent a lot of effort on this, and it was a lot of fun.
Compared the real-world performance of current approaches to online portfolio selection on a variety of portfolios and trading frequencies. A fun project I helped bootstrap and research by developing a Python framework for simulating and backtesting each algorithm.
Collected, stored, and displayed ERCOT capacity and usage data graphs throughout the Texas winter storm crisis.
Prototyped a competitive mental math practice game aimed at preparing for trading interviews. Unfortunately, I never got around to finishing this project. I still think it’s a cool idea, though. Maybe it’ll come back in the future in a multiplayer capacity?
Collaborated with an team of designers and writers to develop a centralized hub for event information.
Developed hybrid redesign of existing Leasing Tablet mobile application using React Native, allowing it to run on iOS and Android. This was my first internship!
Worked on a platform connecting high school students to undergraduate college applications mentors.
This was my first-ever hackathon project! I built an app to compare treatment costs across multiple providers with my hackathon team. We leveraged open hospital chargemasters and Medicare diagnosis codes; since a lot of this data was unclean, I also wrote an ETL pipeline in Python to extract pricing data. Presented at HackRice 2019.
Attempted to repurpose LSTM-based neural sequence-to-sequence language model to the domain of long-form text summarization.
Trained a small machine learning model to detect fraudulent transactions. Learned how to deal with skewed data and metrics such as $F_1$ score.
Launched a web app to help students schedule office hours with their teachers. Replaced the existing experience that was not mobile-friendly. Built with Vue for the frontend and NodeJS/Express for the backend. Launched to peak usage of 100 WAU (weekly active users) at my high school.
Learn how to build a machine learning model to predict Titanic survivors based on the Kaggle dataset.
How people cheat at trivia games, and how the companies running them could stop this from happening.
Developed a question-answering system for HQ Trivia app with realtime OCR in a weekend, with blog post explaining how to mitigate such approaches. Blog post was referenced in articles by The Verge and the Washington Post :)
Explore the depths of your browser history with this introductory data science tutorial.