Mejiro is a financial management tool designed for small organizations, addressing the common challenges of manual bookkeeping and informal price tracking. By combining OCR technology for efficient receipt scanning and GPT-4 AI for data parsing, Mejiro enables users to manage and track their financial records easily. The app allows users to interact with their data using natural language queries for instant insights. It securely stores financial data using MongoDB and offers visualizations through Chart.js, helping users understand their spending patterns and identify savings opportunities. Mejiro aims to simplify and streamline financial oversight for better decision-making.

Makai was developed as part of the Hawaii Annual Code Challenge to address the disorganized reporting system for marine debris at the Center for Marine Debris Research (CMDR). The existing system lacked a centralized database, making it difficult to track and analyze debris trends, with data scattered across Google Drives, Access databases, and Excel spreadsheets. Makai introduced a mobile-friendly platform with a centralized database for debris reports, accessible to both the public and debris removal organizations. This solution enables researchers to monitor debris types and hotspots while streamlining the reporting process. The platform aims to positively impact Hawaii in four areas. First, it contributes to preserving marine ecosystems by facilitating debris removal. Second, it encourages community engagement by raising environmental awareness. Third, it provides data visualizations that empower policymakers, environmental agencies, and researchers to make informed decisions about conservation strategies. Lastly, it helps debris removal organizations optimize resource allocation, making the process more efficient.

This project tackled the challenge of identifying and quantifying diseases and pests in Brazilian Arabica coffee plants using advanced deep learning techniques. The focus was on leveraging the BRACOL dataset, a specialized collection of coffee leaf images, to classify five conditions: healthy, miner, rust, phoma, and cercospora. Early detection and accurate classification of these diseases can significantly improve coffee production by reducing losses and enhancing sustainability. To achieve this, a fine-tuned ResNeXt50_32x4d model was employed, chosen for its robustness in handling complex image classification tasks. Key methodologies included data augmentation to enhance dataset diversity, hyperparameter tuning for improved accuracy, and employing adaptive learning rate strategies with the CosineAnnealingWarmRestarts scheduler. This approach enabled the model to achieve high accuracy while maintaining scalability.

The project aims to analyze sentiments in AI-related discussions on Reddit, employing various machine learning techniques such as conventional methods, Long Short-Term Memory (LSTM) models, and BERT (Bidirectional Encoder Representations from Transformers). Specific goals involve constructing and training models to classify AI-related Reddit content as positive, negative, or neutral. The project compares the performance of different machine learning models, achieving optimal results with BERT and a F1 score of 0.80.

This project developed a Phase Maintenance Inspection Scheduler for military aircraft, optimizing scheduling and tracking during maintenance phases. The system was designed to handle complex scheduling, CSV file imports, event management, and report generation while prioritizing user-friendly design, real-time feedback, and seamless workflows for military personnel.

Department of Education Legislative Tracker (DOELT) is a modern system designed to replace Hawaii's outdated legislative tracker. It improves accessibility and usability with a custom data scraper API that pulls real-time information from the Hawaii State Legislature. The system tracks bills, measures, and hearings, enabling users to search, save, and manage legislative data. Users can also write and track testimony through various roles, from creating initial drafts to final approval. The DOELT ensures the DOE can efficiently evaluate new bills, supporting the mission of enhancing educational excellence and access for Hawaii’s students.

This project collection focuses on various data science techniques applied to real-world datasets. It includes exercises in clustering, time series analysis, machine learning, and data wrangling. Each sub-project provides valuable insights and demonstrates a broad range of data science skills.

Buy From Her is a platform designed to support local female entrepreneurs by promoting their products and services. It helps customers discover and purchase items without extensive travel, making it easier to access products from businesswomen. The project aims to empower women, stimulate local economies, and create employment opportunities while promoting social progress. The design process followed a top-down approach, beginning with research and a mind map, followed by the creation of user flows and prototypes. The platform’s ultimate goal is to foster economic growth and convenience through digital accessibility.

Aeneas Medical is focused on developing a specialized application that assists the H.O.M.E project in effectively managing their medication inventory, supplies, and patient health information. This technology aims to streamline operations and improve accessibility to healthcare services. Hawaii H.O.M.E. Project is a John A. Burns School of Medicine (JABSOM) student-run free clinic staffed by volunteers from the University of Hawaii at Manoa (UHM) and the community. The clinic provides free medical care to the houseless and underserved population of Oahu.

UH Class Critics is a platform for students at the University of Hawaii at Manoa to share reviews of courses and professors. It helps students make informed decisions during course registration by providing insights based on past student experiences. The platform addresses the challenge of limited course feedback by offering a space for students to rate and review courses in various disciplines. Key features include a list of available professors and courses, along with a review system for sharing experiences. This tool enhances transparency, helping students choose courses that best fit their needs.