Building the future
with AI.

Built dual Keras neural networks on a 97,000-record lifestyle and clinical dataset to classify patients across five diabetes stages (No Diabetes, Pre-Diabetes, Type 1, Type 2, Gestational), with separate patient-facing and clinic-facing model variants.

• ▸Patient-side model: 3-layer sequential network (128→64→32, ReLU, BatchNorm, Dropout) trained on lifestyle inputs
• ▸Clinic-side model: trained on 20+ clinical features including HbA1c, fasting glucose, and lipid panel data
• ▸Multi-tab Streamlit clinical system with Patient Dashboard and Clinic Dashboard
• ▸Multi-version model switching via sidebar with confidence score output

An Intelligent Systems 2 course project that uses a diffusion model to transform images of people into their cat counterparts. Demonstrates generative AI techniques through iterative denoising and image-to-image translation.

• ▸Image-to-image translation: human portrait to cat counterpart
• ▸Implementation of forward/reverse diffusion processes

Implementation of a neural network with back-propagation from scratch in C#, demonstrating core understanding of gradient descent, weight updates, and network training without relying on ML frameworks.

• ▸Manual implementation of forward pass, loss computation, and gradient back-propagation
• ▸Built for CS345 Intelligent Systems coursework

An AI-powered Android platform built during the UP Cebu HackaThon 2024 that uses computer vision (YOLO11 + TensorFlow) to detect and classify waste from user-submitted photos, with GPS-based hotspot tagging for environmental monitoring.

• ▸YOLO11 + TensorFlow pipeline for real-time waste detection and classification
• ▸GPS-based hotspot tagging for mapping waste-heavy areas
• ▸Firebase Auth and real-time cloud backend for scalable data sync
• ▸Built under hackathon time constraints with a team of 5

An interactive desktop application visualizing 7 data structures (Arrays, Linked Lists, Stacks, Queues, Deques, BST, HashTable) through step-by-step animations, with an OpenAI-powered chatbot for context-aware explanations.

• ▸OpenAI API chatbot reads live visualization state for context-aware explanations
• ▸Observer, Factory, and Singleton patterns with multithreading for concurrent animations
• ▸Gamified 'Conquest' mode with XP rewards and locked challenge levels for ~50 student users
• ▸Led a team of 5 as Lead Developer

A full-featured LMS serving students and instructors across 5 modules: course management, AI chatbot, calendar scheduler, quiz engine, and task tracker. Built with 266+ commits demonstrating sustained iterative development.

• ▸Singleton, Builder, and Adapter design patterns for scalable architecture
• ▸GSON-based lightweight JSON persistence in place of database dependency
• ▸Polished UI with custom SceneBuilder layouts across 266+ commits

A custom programming language implementation built in C++ for a Programming Language course. Includes lexical analysis, parsing, and language evaluation components.

• ▸Custom lexer and parser implementation
• ▸Programming Language class submission requirement

Pixel-level image processing operations in C# including 3x3 convolution (Gaussian blur, sharpening, edge detection via Laplacian), affine rotation with inverse mapping, and grayscale histogram analysis using direct bitmap memory access via unsafe pointer arithmetic.

• ▸Chroma key background removal using Euclidean color-space distance with adjustable threshold
• ▸Real-time webcam feed processing applying convolution filters live using AForge.Video
• ▸Direct bitmap memory access via unsafe pointer arithmetic for performance