Projects

Bana - Banana Freshness Classification App

Developed an end-to-end ML web app entitled Bana to predict banana freshness. Trained a convolutional neural network on banana images.

Used Python, Flask, TensorFlow, Keras, HTML, CSS, NumPy, and Matplotlib.

View Project

ESRB Rating Classification

Sought to classify video games into ESRB rating categories based on content descriptors.

Trained and evaluated XGBoost, neural network, KNN, C5.0, SVM, KMeans, Naïve Bayes, and CART models for binary and multi-rating classification.

Performed feature selection based on impurity reduction in the XGBoost model.

View Project

LifeSkills Software AI Insights Portal

Built an SDK that can be used by game developers to send data and session info from LifeSkills Software games to the frontend container.

Developed a backend to facilitate the flow of gameplay data from the frontend container to the ChatGPT API for insights generation.

Created an insights portal that allows teachers to query ChatGPT-generated insights based on students, sessions, and games, enabling instructors to tailor educational plans based on student game performance.

Investigation of Discount Factor Effect on PPO Performance on Mario

Trained Proximal Policy Optimization (PPO) algorithms on Mario game frames.

Investigated the effect of several discount factor values on PPO performance based on average reward obtained.

Utilized gym-super-mario-bros and Stable-Baselines3 packages.

Predicting Length of Stay in the ICU Using Lab Test Results, Condition Codes, and Demographic Data

Researched the efficacy of lab test results, condition codes, and demographic data to predict length of stay in the ICU using 41,092 patient admissions from the MIMIC-III dataset.

Engineered features from raw lab test and ICD9 diagnosis data.

Trained an XGBoost regression model to predict length of ICU stay, optimizing hyperparameters via grid search cross-validation.

Sign of the Times: Sentiment Analysis on Historical Text and the Implications of Language Evolution

My research chapter entitled "Sign of the Times: Sentiment Analysis on Historical Text and the Implications of Language Evolution" is published in the international research journal "Advanced Practical Approaches to Web Mining Techniques and Application"

Used R and tidyverse, readr, tidytext, zoo, Hmisc, sentimentr, SnowballC, ggplot2, tm, and dplyr packages.

View Paper

Soundle - An Immersive Game to Improve Mental Health

Developed an interactive, Wordle inspired game that exposes the user to a multi-sensory environment. Collaborated in an Agile team of seven employing Scrum methodology. Implemented several user stories over the course of two-week sprints.

Used React and Python.

View Project

Personal Portfolio Website

Created this personal website!

Used HTML and CSS with Bootstrap.

View Project

Heart Attack Risk Prediction Web Application

Trained Support Vector Machine on heart health data and integrated model into a web application using Streamlit.

Used Python and pandas, Matplotlib, sklearn, pickle, and Streamlit libraries.

View Project

Miscellaneous Mini Projects

I've worked on several miscellaneous Data Science/ML mini projects, such as implementing the K-means clustering algorithm for bean classification to programming logistic regression with MLE from scratch. You can find them on my GitHub!

View Projects