Featured Systems & Engineering

Projects

Full-stack platforms, Retrieval-Augmented Generation (RAG) architectures, and civic solutions engineered with modern AI models, vector stores, and real-time backend systems.

Filter by Domain:(5 of 5)
Undergraduate ThesisAI & Machine Learning

Dual-Branch Digital Image Forensics and Synthesis Detection System

Undergraduate Thesis Project · Multi-Modal Forgery Detection & Forensic Reporting

Developed as our undergraduate thesis project, this integrated forensic image authentication platform is engineered for multi-modal forgery detection, tamper localization, and automated forensic report generation. Implements a dual-branch neural architecture with a multi-gate evaluation matrix to detect and evaluate authenticity across three canonical classifications: Authentic, Spliced, and AI-Generated / Deepfake.

Key Capabilities & Architecture

•Researched and engineered as our undergraduate thesis project, delivering an end-to-end forensic authentication system with courtroom-ready PDF certification
•Dual-branch forensic engine coupling spatial, FFT frequency, DWT wavelet, and OpenCV YuNet facial analysis with TruFor pixel-level tampering localization
•Dense anomaly segmentation decoder utilizing Noiseprint++ sensor noise residuals and MiT-B2 hierarchical vision transformer backbone
•Multi-gate decision matrix fusing global synthesis confidence with localized anomaly probabilities to render 40% opacity JET thermal heatmaps
•Cross-platform client ecosystem featuring a Next.js 16/React 19 web app and secure Tauri v2 (Rust) desktop shell with dynamic @react-pdf/renderer report export
•Asynchronous FastAPI backend with multi-threaded inference queues, JWT role-based access control (Admin & Investigator), and hosted Neon PostgreSQL audit persistence
Stack:PyTorchTruForFastAPIPythonNext.js 16React 19Tauri v2TypeScriptPostgreSQLOpenCVTailwind CSS
GitHub
RAG & LLMsAI & Machine Learning

Ephnyr AI

Full-Stack Multi-Tenant RAG Platform with Groq LPU & pgvector

A high-performance Retrieval-Augmented Generation (RAG) platform designed for document similarity grounding and low-latency AI conversations. Features isolated Knowledge Pods, automated parsing for multi-format documents (PDF, DOCX, TXT, MD), and zero-trace ephemeral messaging.

Key Capabilities & Architecture

•Decoupled Next.js 14 App Router frontend with asynchronous FastAPI backend and LangChain text chunking
•Local CPU FastEmbed (384-dim dense vectors) coupled with PostgreSQL pgvector HNSW cosine similarity search
•Sub-500ms grounded AI inference powered by Groq LPU (openai/gpt-oss-20b) with verified document citations
•Zero-trace session privacy model with multi-tenant storage quotas and room-level access enforcement
Stack:Next.js 14FastAPIPythonLangChainGroq LPUpgvectorSupabaseDocker
Live DemoGitHub
SIKAPTala 2026 HackathonCivic & HealthTech

CheckMatePH

Political Fact-Checking & Civic Engagement Platform

A web-based civic engagement social platform built for Filipino voters to combat misinformation and encourage responsible discourse. Combines automated AI-driven fact checking with expert verification workflows and live civic debate rooms.

Key Capabilities & Architecture

•Built for SIKAPTala 2026 National CS & IT Hackathon (Team Debuggerinas · MSEUF)
•Pre-publication AI fact-checking pipeline evaluating claims against verified public records
•Interactive real-time debate rooms and structured audience voting powered by Supabase Realtime
•Verified politician profiles featuring public claim tracking, integrity scores, and promise fulfillment metrics
Stack:Next.jsTypeScriptTailwind CSSshadcn/uiSupabaseOpenAI APIVercel
Live DemoGitHub
EdTech & RAGAI & Machine Learning

FocusFolio

AI Study Workspace & Automated Document Quiz Generator

An AI-powered study platform that transforms static PDF documents into interactive learning environments. Employs context-restricted RAG pipelines to eliminate distractions and generate dynamic quizzes and flashcards directly from course materials.

Key Capabilities & Architecture

•Strict context-restricted AI assistant that declines unrelated queries to enforce focused learning
•Dynamic generation of interactive multiple-choice exams and front-to-back flashcards from uploaded PDFs
•Split-screen workspace pairing a native PDF reader side-by-side with AI chat and study tools
•Pinecone vector database integration with LangChain document chunking and Supabase storage
Stack:Next.jsFastAPIPythonLangChainPineconeSupabaseClerkTailwind CSS
Live DemoGitHub
Surveillance & HealthTechCivic & HealthTech

BiteTrack

Rabies Surveillance & Animal Bite Case Management System

A healthcare surveillance web application engineered and built for a friend's thesis project. As the full-stack developer, I implemented the system to streamline recording patient rabies cases, tracking animal bite incidents, monitoring vaccination compliance schedules, and automating monthly health department reports.

Key Capabilities & Architecture

•Developed as the primary full-stack engineer implementing a peer's thesis research concept
•End-to-end patient demographic, bite circumstance, and biting animal status registry
•Interactive case database table with real-time patient filtering and dynamic status tracking
•Multi-dose vaccination schedule management with compliance remark calculations
•Asynchronous Supabase database sync with LocalStorage offline fallback and automated Excel export
Stack:Next.js 16TypeScriptTailwind CSSSupabaseRow Level SecuritySheetJS
Live DemoGitHub