It was designed when there was a big earthquake in Bogo City and many people needed therever to quickly find an information source concerning earthquakes, especially aftershocks. Vue.js application to track earthquake movements in the Philippines using an interactive map with Leaflet and marker from OpenStreetMap, with real-time data via web scraping PHIVOLCS dashboard. It includes a sidebar list of recent earthquakes (with magnitude, location, date and depth facets), and a summary panel with statistics for the day and week, as well as for the month.

Designed to give rice growers a quick response when they suspect an issue in the crop, rather than waiting for the agricultural expert. The AI image classification model, trained using the images of rice leaves, classifies eight rice conditions: Bacterial Leaf Blight, Brown Spot, Leaf Blast, Leaf Scald, Narrow Brown Leaf Spot, Sheath Blight, Rice Hispa infestation, and Healthy Rice Leaf, and is served using a FastAPI backend. The React interface allows a farmer to upload or take a picture of a leaf and would get instant feedback; a diagnosis with a confidence score, confidence breakdown for prediction for all classes and actionable Agriculture advice – on what the disease is, how to prevent it, which treatments to use.

My thesis, a high-end career development ecosystem based on AI technologies that closes the gap between graduate and employer. It includes an automated document OCR feature for transcript and business permit verification (using Tesseract.js); it includes an interactive AI mock interview option with real time evaluation of speech by the AI (including theGemini API, for converting speech to text and Deepgram for generating audible responses via speech to text); it has a 24/7 AI generated assessment, and mock interview questions; it has a dynamic resume builder with automated PDF generation; a job suitability score calculated by the AI, automating the smart matching between job seekers and job posts; it is now incorporating verified skill badges and certificates earned from standardized assessments and mock interviews; and a role-based dashboard for both job seekers and employer and admin users. Real-time communication is supported through Socket.io, automated pipelines for AI generation with Heroku Workers, email delivery with Mailtrap, and real-time alerting with Web Push Notifications.

My first fully implemented full stack project that ended up in production while learning full stack. ApplyTrack is a software for anyone who has ever had trouble finding a job in an unorganized spreadsheet application. Made with MERN, React as the front end, serving an Express and MongoDB server on Render which enclose an excel sheet, but correctly into a dashboard. I implemented the handlers and controllers that handle each route; separated Mongoose models, route controllers, and views to make codebase easy to navigate as features were added on; used an off-the-shelf token signing service (JWT) to ensure the receiving application on the server has it, and create a protected route middleware that filters it and only allows the subscribed user to access the dashboard; and instead of using an off-the-shelf provider for authentication, I built my own JWT-based sign and verify authentication system from scratch, which I passed through the session handling procedure to the front. On top of that is full CRUD atop applications and a summary panel being thrown which has the total number of applications, success rate and offers received. A pipeline view filters all of your applications by status – Pending, Interview, Offer, Accepted, Declined and Ghosted – and provides you a distribution chart and an activity graph for viewing application progression by time.

An intelligent, machine learning-powered web application designed to assess and predict student dropout risks based on academic, lifestyle, and demographic factors. It features a modern, interactive React frontend and a fast API backend. The system provides real-time percentage predictions and visually distinct risk levels Low, Medium, High alongside actionable, AI-recommended interventions tailored to the student's specific risk category. Powered by a pre-trained Logistic Regression model served via FastAPI.