HAO • PORTFOLIO
From Passive Monitoring to Proactive Security Defense
AI & Cybersecurity
Developing advanced ML models for spam detection, real-time intrusion detection systems (IDS), and fraud heuristics.
Cloud Infrastructure
Cloud Infrastructure: Designing and deploying secure, scalable, and resilient cloud-native systems using Docker, CI/CD, and DevSecOps practices.
Network Communications
Implementing robust server setups, edge computing integrations, and custom telemetry data pipelines.
I am LOI CHIANG HAO
Cybersecurity & AI Engineer | Cloud Infrastructure | DevSecOps
Biography
I am a final-year Information Technology student at Universiti Tunku Abdul Rahman (UTAR), specializing in Communications and Networking, with a focus on cybersecurity, AI systems, and cloud infrastructure. I am particularly interested in how systems can evolve from passive monitoring to proactive defence through secure design, intelligent detection, and resilient infrastructure. Through hands-on projects, I explore security engineering, AI-powered systems, distributed architectures, and DevSecOps practices. My goal is to build practical systems that remain secure, scalable, and dependable under real-world operational challenges.
Core Skills & Specializations
Experience & Education
My academic and professional journey
IT Support Specialist (Internship)
Meteorsoft Sdn Bhd
Nov 2025 — Jan 2026
Managed system deployments, customized business reports using DevExpress, supported server setups, and configured networking hardware. Provided direct on-site and remote client support for the Auto Power Suite (APS) ecosystem.
Undergraduate IT Student
Universiti Tunku Abdul Rahman (UTAR)
Oct 2022 — Present
Final year student (MUET Band: 4.0). Spearheading research in AI-powered Email Threat Detection systems and gaining hands-on expertise in digital forensics, ethical hacking, internetworking, and operating systems.
Featured Projects
Systems I've designed and built
AI Model Atlas: Cognitive RAG Architecture Simulator
AI Model Atlas is an open-source learning platform and dual-track reference architecture designed to help developers build secure, enterprise-grade AI systems beyond traditional “Chat-with-PDF” wrappers. Combining a bilingual curriculum with a production-ready sandbox, it demonstrates advanced architectures like GraphRAG, hybrid retrieval, and automated CRAG web fallbacks. Crucially, it introduces a complete AI Security Engineering blueprint that includes adversarial red teaming, prompt-injection defences, and runtime guardrails. AI Model Atlas serves as both an educational roadmap and a practical blueprint for building intelligent, attack-resilient AI applications.
Day 396 • A Cybernetic Experiment in Oblivion
Designed and engineered an anonymous digital time capsule that combines immersive cyberpunk storytelling with production-grade system architecture. Built on a CQRS-inspired, append-only event model using Supabase and PostgreSQL, the platform leverages strict Row-Level Security (RLS) policies to enable anonymous submissions while enforcing immutable data access patterns. To overcome regional network restrictions, I architected a dual-line Caddy reverse proxy within a Docker container and implemented intelligent client-side endpoint routing for reliable cross-region API communication. Additional security hardening, automated sanitization, and archival pipelines ensure long-term data integrity without relying on traditional backend services.
HAO • PORTFOLIO (AI-Powered Portfolio Platform)
HAO • PORTFOLIO is a production-grade AI-powered engineering platform built with Next.js 16, TypeScript, Tailwind CSS v4, Sanity CMS, and Vercel Edge Runtime. It showcases real-world full-stack system design through a secure, high-performance architecture featuring HaoGPT, a multilingual RAG-based AI assistant with structured retrieval, prompt-injection protection, and dynamic context routing. The platform implements a lightweight zero-trust backend-for-frontend (BFF) architecture, Redis-based rate limiting, signed asset delivery, audit logging, and CMS event tracking. Optimised for global edge deployment with sub-100ms latency and ISR-driven content delivery.
Cloud-Based Distributed E-commerce Analytics Platform
A high-performance distributed e-commerce analytics platform built with Python, Streamlit, and Flask, featuring real-time data sharding across geographic nodes. Designed with a robust asynchronous aggregation engine (httpx + asyncio) and intelligent request routing, it ensures high availability and partial fault tolerance while seamlessly integrating with Supabase for cloud-native PostgreSQL data management.
Kedai Kopi Gedok Website & Management
A highly secure, full-stack landing page and admin system for a local F&B business. Engineered a zero-trust Backend-for-Frontend (BFF) architecture featuring stateful session revocation, global distributed rate-limiting, and AI-driven spam honeypots to neutralize CSRF, XSS, and bot attacks.
AI-Powered Email Threat Detection Extension
Developed an AI defensive Chrome extension using a Multi-Layer Perceptron (MLP) and One-Class SVM. Fused Content Metrics with Network Forensics (SMTP/SPF) to preempt zero-day phishing mutations. Implemented SHAP for advanced Explainable AI.
TrustGuard AI Fraud Detection
Built for the V HACK 2026 Hackathon. A hybrid fraud detection platform combining XGBoost ML models and Behavioral Heuristics to secure digital wallets. Features geospatial monitoring and neural network relationship graphing.
Smart Factory Simulation & IIoT Dashboard
Designed a comprehensive Smart Factory Simulation using Node-RED to replicate manufacturing lifecycles with real-time logic control, inventory management, SVG interactive dashboards, and IIoT MQTT telemetry.
Certifications
Continuous learning & validation
Generative AI/LLM Concepts
Alibaba Cloud
ISC2 Candidate
ISC2
Gemini for Data Scientists and Analysts
Connect and Protect: Network Security
Google Cloud
Responsible AI for Developers
Google Cloud
Level 1 TRIZ Practitioner
International Triz Association
Cybersecurity Job Simulation
Mastercard & Datacom