Year 2 Computer Science · Multimedia University · Kuala Lumpur
I'm Micheal — a second-year Computer Science student at Multimedia University, majoring in Software Engineering.
I'm someone who's never been content just using things. As a kid I took apart whatever was in front of me; later that turned into building software, fixing hardware as an IT technician through college, and getting unreasonably curious about how anything works under the hood. The shape of the work keeps changing — the curiosity is the constant.
I'm a GitHub Campus Expert, and I directed BarCamp Cyberjaya 2025 — MMU's largest student-led tech event.
A quick hello — who I am and what I'm working on.
Agentic AI · AIBC
AIBC's evolving agent platform: DCS (smart RAG) → DAC (knowledge-graph + vector RAG) → DACv2 (autonomous, Medusa-powered). The aim is to turn unstructured documents into something an agent can reason and act over.
Reusable agentic engine · AIBC
AIBC's reusable agentic engine, which I co-founded. Security gateway in front of every tool call, MCP-native transport, and a hybrid neuro-symbolic cascade that combines cheap heuristics with multi-LLM validation before any irreversible action. Powers DACv2; the first prototype task built on top of it is the Unsubscribing Agent, used to stress-test the engine's autonomy.
Smart RAG · v1 of DAC
The first milestone of the DAC line at AIBC. AI document processing pipeline — Chandra OCR (deployed on AWS SageMaker) extracts text from scanned PDFs and images, structured chunks land in PostgreSQL with pgvector + pgai for hybrid semantic + keyword retrieval, and a LangGraph RAG agent answers queries with persistent long-term memory.
WhatsApp AI sales agent
A WhatsApp AI sales agent for property agents — replies to leads, follows up, and books viewings while you're asleep, driving, or running errands. Built originally for a Malaysian property agent: property-listing RAG, knowledge-base RAG, human-takeover dashboard, and pluggable LLM providers (Anthropic / OpenAI / Google) so the agent can be tuned for cost or quality.
AI-driven emergency alert system
Real-time disaster alert and search-and-rescue platform. Nine AWS Lambda functions (Serverless Framework) scrape Reddit for disaster reports, validate them against NOAA / USGS data, and serve heatmaps and evacuation routes through an MCP-bridged Google Maps integration. Amazon Bedrock Nova Pro does the language work.
Emotion-aware reply assistant for couples
An AI reply assistant built for couples — because the wrong text at the wrong moment is its own kind of disaster. Detects emotion from your partner's message, reads the air, and drafts replies tuned to your intent (calm them down, be supportive, apologise, change the topic) while a ChromaDB-backed memory keeps the conversational context across turns.
Cross-platform meal recommender
A small app for the chronically indecisive. Suggests meals based on user preferences and location.
The cadence is the point — each one a forced exercise in scoping, shipping, and not falling asleep at 3am.
AI finance app for Malaysian gig workers — handles irregular income, self-managed tax, no-employer EPF, and receipt parsing for people the banking system still treats as an edge case. Next.js + Django + Gemini 2.0 Flash + pgvector.
AWS-serverless disaster monitor — scrapes Reddit, runs NLP, cross-validates against NOAA / USGS / NASA, and renders verified events on a live map.
Prediction-market arbitrage scanner — pulls real-time odds from Polymarket and Kalshi side-by-side and flags spreads worth taking.
Insights chatbot for Grab MEX (merchants) under the +AI vision — RAG over transaction data so partner restaurants can ask plain questions of their own numbers. Domain 3, Task 2.
AI-powered healthcare assistant for disease management — built for the AI & Data Science track.
Twitter-sentiment-to-market predictor under the Social Trading challenge — tag the Eliza agent under an influencer post and the dashboard surfaces the trend for users to call bullish or bearish. Eliza ate too many bugs; pivoted to Make.com inside 24 hours to ship. With Sky Yap, Lim Xin Yi, and Mui Rui Xin.
Real-time cybersecurity dashboard — Scikit-learn models on NSL-KDD / CIC-IDS-2017 score incoming logs for phishing, malware, and insider-attack likelihood, with geo-located IPs, anomaly detection, and one-click block actions. Flask + MongoDB + React, WebSockets for live updates. Built for the AI-driven Cybersecurity Solutions track.
Selected for GitHub's competitive student technologist leadership program.
Led delivery of MMU's flagship technology event — 250 attendees, MDEC partnership.
Directed marketing for MMU's largest student-led IT community across 6 major events.