Projects

32 web-based diagnostic tasks for aphasia research, each deployed via Qualtrics with millisecond-precision timing for EEG synchronization.

The tasks cover comprehension, production, repetition, and naming, giving researchers granular data on how language breaks down and recovers. Built and maintained as part of my work at the DARN Lab.

As part of Alycia Laks, MS, CCC-SLP's research paper, I helped create a public version of the task suite available as standalone HTML versions.

These versions don't collect data for our lab, but include on-the-spot data analysis presented directly to the user, making the tasks accessible for broader research and educational purposes.

A command-line tool that reads DNA sequence files and analyzes codon usage, amino acid translation, and sequence statistics.

Built as a side project during my work as a Georgetown Lab RA, this tool evolved into a fairly complete bioinformatics resource.

VR environments for EEG data collection on Oculus Quest 2. The hard part was getting audiovisual stimuli to fire precisely enough to be meaningful in a neuroscience context, where timing tolerance is extremely small.

Includes customizable experiment protocols and network-trigger integration with external EEG systems.

A minimal, hand-crafted portfolio site built from scratch without frameworks. Designed for clarity and fast load times.

The site focuses on showcasing work and writing with a clean, typographic approach. No JavaScript frameworks or unnecessary dependencies.

Automated unit testing for C++ assignments in Georgetown's CS department. It runs student submissions against a comprehensive test suite and returns detailed feedback.

Built partly out of necessity. Manual grading at scale simply does not hold up.

A study tool that takes uploaded notes and generates flashcards, quizzes, and summaries using the Gemini API.

I handled the backend: storage, user registry, and most of the code review and merging across the team.

A tool that automates testing by simulating user input from files. It reads pre-defined input files and passes them to a target program, eliminating the need to manually type inputs for each test.

Built to streamline the testing and grading workflow. Includes a C++ program for file reading, Bash scripts for automation, and a flexible folder structure for different testing scenarios.

Built for Georgetown COSC 3590 (Data Mining), this project runs neural-network experiments over the Adult ARFF dataset through two pipelines: a custom C++ implementation and a Python scikit-learn MLP baseline.

Both runners generate timestamped artifacts under results/ to enable direct output comparison across experiment configurations.

A Naive Bayes processing pipeline for Adult ARFF data built for COSC 3590. The project includes configurable discretization, fold-level outputs, and evaluation summaries saved in timestamped results directories.

It also includes an easy-mode workflow for quick execution and testing.

A utility project for processing ARFF files (non-sparse ARFF), created to support the neural-network_processor and naive-bayes_processor workflows.

This parser/tooling layer helps standardize data loading and preprocessing across both class projects.

A deep learning model for generating audio from text prompts, built as a final project for my deep learning course.

More of an experiment than a polished product, but the model produced usable outputs and helped validate the full training and inference pipeline.

A database-backed application with a front-end UI, similar to an Access-style workflow but built from scratch.

Final project for Georgetown's Intro to Databases course, built with Berk Ugurdag. We designed the schema, built the MySQL backend, and wired up a functional interface on top of it.