Languages

Frontend & Frameworks

Backend & Infrastructure

Tools & Creative Suite

AI & Machine Learning

A desktop AI assistant that's invisible to screen capture software. It transcribes speech in real-time with sub-200ms latency and uses local RAG to search your personal knowledge base privately.

INVISIBLE Custom Win32 hooks render on a separate desktop layer, undetectable by screen capture
SUB-200MS Real-time speech-to-text with voice activity detection
LOCAL RAG Private vector search over resume, project docs, and knowledge base

Fine-tuned Llama 3.1 8B using knowledge distillation (Gemini 2.0 as teacher) to analyze Terms of Service agreements clause-by-clause. Detects predatory practices and legal risks. Runs entirely on consumer hardware with $0 infrastructure cost.

40% reduction in false positives vs baseline | 2x faster training with Unsloth | Near-GPT-4 quality on laptop

Reverse-engineered Meta's encrypted Thread Anchored Messaging (E2EE) protocol. Custom-patched Messenger APK with Ghost Mode, anti-typing indicators, and unsend recovery. Works on non-rooted devices. Featured on XDA Forums.

GHOST MODE Read messages without "Seen" receipts, works in standard AND encrypted chats
NO TYPING Compose without the "Typing..." bubble
UNSEND RECOVERY Locally intercepts and logs messages before they vanish

Production booking platform I solo-developed from scratch. Includes a RAG-powered AI chatbot for customer inquiries, an admin analytics dashboard with real-time KPIs, and automated booking conflict detection. Live and handling real traffic.

~60% reduction in booking errors | ~70% of inquiries handled by AI | ~50% faster response time

Award-winning Android app that modernizes livestock management. NFC/RFID ear tag scanning, Firebase real-time sync, offline-first architecture with automatic sync, and team-based role management. Took data retrieval from minutes to seconds.

_{19th Business Idea and Development Award | Philippine Chamber of Commerce and Industry}

Java desktop app that talks directly to the game's internal LCU API through real-time WebSocket connections. Auto-discovers running game processes, extracts auth tokens, and reacts to game state changes in milliseconds.

AUTO-ACCEPT INSTA-LOCK BENCH SNIPER BULK DISENCHANT via real-time WebSocket state machine

Facility management web app built for Saint Louis University (SAMCIS). QR-based check-in with time-limited codes, conflict detection that prevents double-bookings under load, and domain-restricted OAuth (@slu.edu.ph only).

15 min to 30 sec booking time reduction | Zero double-bookings | QR check-in with auto no-show handling