HouseBot
Virtual Rosetta
Count BOTula
End-to-end framework for a robot designed to tidy a house. Incorporates a Convolutional Neural Network (CNN), Region Proposal Network (RPN), and additional feature mapping to detect objects in a room scene. The object labels are fed into ConceptNet that maps a list of verbs as the actions the robot should perform on a specified object to the object noun. Trained on scraped
AI2-THOR room scenes.
Python, Tensorflow
November 2021.
Virtual reality environment named "The Virtual Rosetta," developed in Unity. Ancient Vietnamese drawings were organized according to a hierarchical clustering with BERT embeddings. Evaluated the effectiveness of several user interface designs before implementing the final design. Workshop paper published and presented at the
6th IEEE VIS4DH Conference on October 24, 2021.
Python, Unity, Word2vec, Scikit-learn
October 2021.
Adversarial search algorithm with a Voronoi + Dijkstra heuristic to compete against other bots in the game of Tron. Generated dataset comprised of board game configurations and corresponding winners of existing bots. Trained a supervised regression model that predicts the probability a player will win the game based on a given board game configuration.
Python
December 2020.
LanguBrain
Upserve To-Go
UScreen
A translation application utilizing Apple's Speech Recognition and Speech Synthesizer frameworks and the Google Translate API. User can translate between 34 languages by speech or text. Detects languages being spoken without the user having to specify the origin language.
Available on the App Store.
Swift
July 2019.
Mobile application for restaurant management technology company Upserve. Notifies user of nearby Upserve restaurants through location-based notifications and proposes recommendations and rewards based on user preferences. Utilizes Apple's MapKit and CoreLocation as well as Firestore, Firebase Storage, and Firebase Authentication.
Swift
August 2019.
An iOS application utilizing Apple's ARKit to display the parts of the user's face most susceptible to sun damage. Used OpenWeather API to create an in-app weather feature along with Firebase to store user data.
Swift
April 2019.
Villar
Vagrant Coffee
Model Selection
Real estate filtering web application. Users are able to filter nearby houses through location, property type, and listing. Sort by price as well as receive current weather data utilizing the OpenWeather API.
JavaScript, React
December 2019.
Complete redesign of the coffee cart startup Vagrant Coffee mobile application. Conducted user testing and gathered metrics for task completion, including error rate, time on task, and success rate.
Adobe XD, Balsamiq
November 2019.
Binary logistic regression model with Tikhonov regularization used to classify a cell as malignant or benign. Trained on the UCI Breast Cancer Wisconsin (Diagnostic) dataset.
Python
March 2019.
Sentiment Analysis
ChatBee
Musical Style Transfer
Hypothesis testing to determine the sentiment of conservative versus liberal news articles. Performed multivariable regression using StatsModels OLS and accounted for bias present in several news sources.
Python
May 2020.
Chatroom web application in which the user can join a chatroom, post messages, and view other users' messages. Utilizes MongoDB for backend.
JavaScript, JQuery
March 2019.
A Sequence to Sequence Recurrent Neural Network (RNN) that transfers the style of famous classical music pieces into jazz. Trained on the Piano Dataset of 349 classical pieces and 349 jazz pieces.
Python, Tensorflow, PrettyMIDI
October 2019.
