Projects
RoutineRemind
RoutineRemind is a mobile (iOS/Android) and web application that was selected by U.S. Representative Jennifer Wexton as the winner of the Congressional App Challenge from Virginia's 10th District. Since then, it has been provisionally patented and is currently utility patent-pending. The application is designed to aid individuals with cognitive and speech impairments, with a focus on children with autism and their families. Using the app, parents can create daily schedules for their children by recording multiple action items for the day. When a user asks a question about their routine, the app plays back the matching recording from the parent, which reduces maladaptive behavior in children with autism. The app incorporates machine learning algorithms and natural language processing libraries to match the child's question with the corresponding recording from the parent. The app's frontend was built using Javascript, Typescript, and HTML/CSS. Authentication and scheduling data in the backend was stored in the backend through Google Firebase.
Media
RoutineRemind wins Rep. Jennifer Wexton’s 2022 Congressional App Challenge in Virginia’s 10th District
TJ students win local Congressional contest by designing app to help kids with autism
Loudoun Students Win 10th District Congressional App Challenge
TJHSST students win 2022 Congressional App Challenge
& more
EyeLS
Neurodegenerative disorders, characterized by progressive degeneration and loss of function in the nervous system, are the leading cause of physical disability worldwide. In particular, neurological conditions such as amyotrophic lateral sclerosis (ALS) impede patients’ ability to perform voluntary muscle movements such as walking or speaking. Accordingly, researchers have attempted to develop augmentative and alternative communication (AAC) mechanisms and speech-generating devices (SGDs) to facilitate communication for individuals with speech or language impairments. However, the advantages of AACs and SGDs fail to overcome their limitations due to the inadequate availability and affordability of these devices. Conversely, gaze-tracking technology offers a promising solution to enhance non-verbal communication, but current models are both variable and imprecise. EyeLS utilizes JavaScript, HTML, and CSS while incorporating Visual Studio Code as its integrated development environment (IDE). The application’s webcam gaze-tracking implements a ridge regression model to map multi-dimensional eye feature vectors through click locations as training points. Kalman filtering, a Monte Carlo approach to linear quadratic estimation, processes and refines gaze vectors to maximize EyeLS’s precision. Accuracy was calculated using the Euclidean distance between the model’s predicted gaze estimation and click location. The average prediction error made by the ridge regression model over 1500 data points was xÌ„ = 88.4 pixels with s = 26.9 pixels, correlating to a mean calibration accuracy of 92.23%. EyeLS is a propitious step forward in overcoming the limitations of modern gaze-tracking technology through its precise calibration algorithms, providing an accurate and accessible means of communication for patients with neurodegenerative disorders.
Memory Lane
Memory Lane is a mobile application designed to provide people with Alzheimer's and other memory loss-related conditions with a platform to replay memories from their past. The app requires a login through email for each user, and upon authentication, the user is presented with three options: Add an entry, Ask a question, and a unique feature called "A Memory A Day".
In the “Add an entry” section, users can record a significant event at the time of its occurrence, summarize it with a keyword or phrase, and attach an associated image. The recording is then saved, and the key phrase is used to identify the necessary output in the future when a question is asked. In the Ask a question section, users can ask the app a question about a specific event they have forgotten. Using an NLP model built on third-party libraries such as RASA and Tensorflow, the app will then search for a keyword match in the question asked and play the associated audio recording of the event. The app will also display the associated image, helping the user remember the event in greater detail. The "Memory A Day'' feature is designed to remind users of a random memory each day, helping them keep their minds active and engaged. This feature not only provides users with an opportunity to reminisce about their past but also helps them build and maintain social connections, which can have a positive impact on their overall well-being. Overall, Memory Lane's intuitive design and features enable users to maintain their connections to their past, promoting a sense of well-being and quality of life.
LymeML
LymeML is an application that implements novel machine learning algorithms, including convolutional neural networks (CNNs), to diagnose Lyme disease. Contemporary methods require a blood test and waiting for three to four weeks on average, during which time symptoms worsen detrimentally. However, LymeML offers accurate results in just a few seconds. This project was featured as the Grand Prize winner in the AI for Humanity Hackathon in 2024 and won the MecSimCalc Week Long Build-A-Thon in 2022.
Disease Detector
Disease Detector is a web application that features machine learning classification and a self-check symptoms diagnosis to recognize and distinguish a plethora of diseases with accurate, immediate results. This project aims to aid individuals in need of a preliminary diagnosis and is beneficial in remote environments without significant medical attention. Disease Detector was featured as the Red Cross Challenge winner in HackTJ 9.0, America's largest high-school run hackathon.