×

AutoAccred: Automated Document Classification and Management for School Accreditation

This project served as our thesis title for our coursework, completed and defended in 2026. It focused on automating document management, primarily for school university accreditation processes. Current manual processes are time-consuming and prone to errors, especially with large volumes of documents. AutoAccred aimed to streamline and improve the accuracy of these processes, by integrating a machine learning-based document classification system with an automated document management system.

The document classification prediction architecture consists of three main components: a document classification module, named entity recognition (NER) module, and a keyword analysis module. The document classification module uses a fine-tuned BERT model to categorize documents into 279 predefined classes relevant to school accreditation. The NER module extracts key entities from the documents, such as dates, names, and specific accreditation-related terms which it uses to adjust the confidence scores. Finally, the keyword analysis module uses TF-IDF to identify important keywords in the documents, further refining the classification results. Predictions are then pushed to users. From over 2000 folders, relevant ones are recommended based on the document classification, together with each folders' descriptions. Selected folders are then given a copy of the original input, and are saved on the cloud storage for easy access and retrieval later.

The models was trained and evaluated using a synthetic dataset of over 33,000 samples. The combined system achieved an overall accuracy of around 96%, with precision exceeding 97%. This high level of performance demonstrates the effectiveness of integrating multiple machine learning techniques for document classification and management.

I lead the group that developed this system, overseeing the integration of the various machine learning components and ensuring the overall functionality and performance met our project goals. I was responsible for designing the system architecture, coordinating the development efforts, fine-tuning the two models, and conducting extensive testing to validate the system's accuracy and reliability. I also designed the backend for the system, as well as the software used in creating the synthetic dataset for model training and evaluation.

Technologies used in this project include React and HTML5 for the frontend development, providing a user-friendly interface for document upload and management. The backend was developed using Python, leveraging the HuggingFace Transformers library for implementing the BERT models, and pytesseract for optical character recognition (OCR) to extract text from scanned documents. SupaBase/PostgreSQL was used for database management, ensuring efficient storage and retrieval of document metadata and classification results. CloudFlare tunneling was employed to facilitate secure web migration and access to the system.

System Architecture

×

CCS Information and Room Tracking System

CCS Information and Room Tracking System is a final project made for the subject Software Enginnering, completed and defended in 2025. It focused on organizing and automating various administrative tasks within the College of Computer Science, enhancing efficiency, security and accountability. It also automates the generation of required reports for the college such as attendace reports, room usage reports, and daily time records.

The system integrated an IoT system and a web system to deliver a comprehensive migration to digital processes within the college. It aimed to centralize all information arount the college, including student and faculty attendance, schedules, room and PC usage, events, and student and faculty personal information. By integrating an RFID system, the system was able to monitor attendance and room usage in real-time, enhancing security and accountability. It also allowed other student or faculty to view where a student or faculty is currently in. The system also features an event organizing module, allowing students, faculties, and administrators alike to view upcoming events and schedules, with email notifications for reminders. Report generation is also automated, allowing administrators and faculties to easily generate attendance reports, room and PC usage reports, and daily time records with just a few clicks.

I designed the wiring schema needed for the IoT integration to the web system of this project. I was also the one responsible for making the pipeline on how all the information around the system would flow, how the system uses these data, the expected outcomes, the security put in place, as well as how the IoT system would interact with the web system. I also made the entire IoT system, including the Arduino code and the wiring needed for the RFID scanners and solenoid locks.

Technologies used in this project include Flask for the backend development, providing a robust framework for handling web requests and database interactions. The frontend was developed using Jinja templating engine along with Bootstrap and HTML/CSS, ensuring a responsive and user-friendly interface. The IoT components were built using ArduinoUNO microcontrollers, RFID cards and scanners for attendance and room tracking, and solenoid locks for automated security. MariaDB was used as the database management system, ensuring efficient storage and retrieval of all college-related information.

System Architecture

IoT Wiring Schema

×

Community Disaster Hub

This project is made in a hackathon called "Blue Hacks", held in Ateneo de Manila University in 2025, where I and seven of my groupmates were chosen to represent the school in this inter-university competition. The theme "Response to Natural Calamity" was only announced at the start of the event.

With the recent calamities that hit Philippines, social media became a place where people posts their current situation, in hopes that rescuers, LGUs, or concerned citizens can help them. Although this works for the time being, there is clearly no centralized system for this kind of response, and there is no assurance that the posts are seen by the right people.

Community Disaster Hub aimed to address this gap by providing everyone with a platform to post their current situation and needs, and for anyone to respond to their aid. The app is split into two modes. First is the posting. People who are in need can post here and describe their current situation. After posting, the user's contact number that they used to register, and their current geolocation will automatically be included in the post. Whenever someone responds to their post, a text will be sent to said number. Second part is the responders' screen. Here, rescuers and civilians alike can view all posts and offer their assistance. By pressing the "Respond" button, responders can tag the post as someone they are currently helping. This "Respond" count lets other rescuers know that someone is already tending to that post, so that they can help other people that has no or few respond count. Victims have the ability to mark their post as resolved when they are done receiving aid, letting rescuers focus on other people who still needs help.

I co-developed the entire frontend and backend of this project. I also co-designed the app design, functionalities, and entire system architecture.

Since this app needs to be used by everyone regardless of their device, a cross-platform application is necessary. Flutter was used to develop this project, and was exported to Android, Windows, and Web platforms. Firebase was used as the database to allow real-time updates of posts and responses.

×

MZQR: An Interactive Zoo Catalog for Manila Zoo

MZQR is an application made in 2024 for a subject "Operating Systems", created to understand the architecture and nuances of different operating systems. As such, cross-platform development is the primary focus of this project. Made with permission from Manila Zoo's director. The application is an interactive zoo catalog for Manila Zoo, utilizing QR code technology to provide visitors with information about the animals in the zoo, reducing the need for large and outdated signages.

The application allows zoo visitors to scan QR codes placed near animal enclosures, which then directs them to detailed information about the animal, including its species, habitat, diet, and interesting facts. The app also includes animal information that is specific to the Manila Zoo, such as the animal's name, years it has been at the zoo, and any unique characteristics associated with it, and year of acquisition. All changes are updated real-time. The administrator of the zoo, as well as the zookeepers have access to the admin side of the app, and can make changes to the details of the animals at anytime. The app also have automated report generation on how many visitors the zoo got by tracking the number of unique device scans per day, month, and year. In the end, MZQR has been deployed to Android, iOS, and Windows devices, as well as to a Web System.

I designed the system architecture of the app, including its functionalities and restrictions. I was also responsible for co-developing the entire frontend and backend of this project, as well as co-designing the app layout and user experience, as well as implementing the real-time database of the app. I led the research team that gathered all the necessary information about the animals in Manila Zoo from the zookeepers, staffs, and administrators, ensuring accuracy and relevance of the content provided in the app.

Like mentioned before, MZQR has been developed using Flutter and Dart as the programming language, allowing for a cross-platform deployment without rewriting the codebase. XCode was also used for importation of the app to iOS platform. Firebase was used as the database management system, enabling real-time updates and easy data retrieval for animal information and visitor statistics. Along with it, Firestore was used for cloud storage of images and other media files related to the animals, the zoo, and the application.

×

SK Projects Management: Integrated Development Framework

Made in collaboration with Sangguniang Kabataan from Tanza 1, Navotas City, made in 2023. With the current agenda of the government to switch all aspects of their system to a digitized version, from the department level to the LGU levels, SK is often overlooked, resulting to them still using traditional and manual methods of handling day-to-day activities. SK Projects Management is a comprehensive system focused on helping SK Kagawads and Chairman perform their duties effiently, by providing them with tools to manage their projects and events. All functionalities of the app was decided by my team and the SK themselves through regular online meetings.

The system is a plethora of sub-system compiled together, covering a large part of SK's daily activities. First is the KK Profiling. Here, SK can input the information about youth in the Barangay. Their data is used for the other functionalities of this system. Second is the Event Management System. Here, SK can organize events and schedules through the use of a built-in calendar. Automatic email notifications are pushed on the day before the event as reminder about the upcoming event. Third is the Budget Tracking System. Here, SK can monitor their budget and expenses, with built-in graph generation for trends and analysis. They can tie this to the Event Management to manage their event budgets, and include this specificity in the generated graphs. Lastly is the Sports League Information System. Here, SK can manage sports leagues in the Barangay, including team information, schedules, and scores. The statistics of the players are automatically tallied for easy tracking. All of these functionalities are integrated together, making this system a one-stop-shop for SK Kagawads and Chairman to manage their day-to-day activities.

I was the lead developer of this project, responsible for designing the overall system architecture and ensuring the seamless integration of all sub-systems. I also co-developed the backend of the application, implementing the core functionalities and database interactions. Additionally, I collaborated closely with the SK officials to gather requirements, provide training, and ensure that the system met their specific needs and expectations.

The system was developed using Visual Basic .NET for the application development, providing a robust and user-friendly interface for SK officials. SQLite was used as the database management system, ensuring efficient storage and retrieval of all data in the system, and making sure that the database is offline-capable since internet connection is not always stable in the area.

CONTACTS

HELLO, I'M ROLAN!

BACK END / SOFTWARE DEVELOPER

PROJECTS

Automated Document Classification and Management for School Accreditation

Community Disaster Hub

CCS Information and Room Tracking System

MZQR: An Interactive Zoo Catalog for Manila Zoo

SK Projects Management: Integrated Development Framework

ABOUT ME

Rolan Jay P. Samonte

SKILLS

Programming Languages

Technologies and Frameworks

Productivity and Creativity Tools