data behave
...then I build & ship
Software Engineer · Shah Alam, Malaysia

I turn raw, messy data into decisions a business can trust, then build thesystems, APIs, and interfaces that put those decisions to work. From the pipeline to the page, I like owning the whole path:data, backend, deployment, and the web layer on top.
Over 20+ years I've gone from speech-recognition research to production data and software engineering, where I learned to measure first and build second. These days a single project might span a data pipeline, the model on top, the API that serves it, and the web app people actually click. Increasingly it also means lifting a legacy system onto Microsoft Azure and keeping it healthy once it lands there.
Lately my favourite teammate is Agentic AI. I build with tools like Claude and Antigravity to turn an idea into a working prototype in hours, and I write the governance and guardrails that let a whole team do the same safely. Fast is easy. Fast and trustworthy is the interesting part.
I'm most at home in Python, C/C++ and SQL, with MATLAB, Go and Java when a problem asks for them. I reach for whatever fits: a web stack one week, Docker, Kubernetes and CI/CD the next. New tools don't scare me, they're the fun part, and if something can be automated, I'll probably try.
01Things I've built
SIRIM Marketplace Platform
One front door for SIRIM's enterprise products and services. I lead the build of a unified marketplace that centralises discovery across business units, so a customer finds the right lab, certification or service in a few clicks instead of a few phone calls.
Legacy Re-platforming on Azure
Critical legacy systems moved onto the Microsoft Azure ecosystem without pausing the business that depends on them. End-to-end deployments, DevOps pipelines, and the architecture decisions that make the next migration easier than the last.
Agentic AI Governance & Guardrails
The rules of the road for AI-assisted development at SIRIM. A governance framework plus technical guardrails, so the team ships fast with agentic tools like Claude and Antigravity while every AI-integrated application stays secure and compliant.
Enterprise Data & AI Platform
The analytics backbone at SIRIM. Automated pipelines feed a central warehouse, ML models serve predictions behind clean APIs, and dashboards turn it all into decisions people trust. Built in Python, containerized, and shipped through CI/CD.
GenAI Knowledge Assistant
A retrieval-augmented assistant over thousands of internal standards and lab documents: semantic search with grounded LLM answers, so staff find what they need in seconds instead of digging through PDFs.
02My journey
03Tools I reach for
How my stack fits together — drag a node to give it a nudge.
- Python
- PyTorch
- TensorFlow
- Scikit-Learn
- NLP
- Statistics
- C / C++
- MATLAB
- Deep Learning
- SQL
- Power BI
- Dataiku
- ElasticSearch
- Kibana
- Docker
- Git
- Linux
- LLMs / RAG
- Agentic AI
- Embedded / IoT
- Go
- Java
- Kafka
- CI/CD
- Kubernetes
- Azure
Certified · IEEE CertifAIEd (2024) · Microsoft Power BI Data Analyst Associate (2025) · Deep Learning Eng. (CERTIFAI) · Graduate Engineer (BEM)
04Data is beautiful
Even this page is a dataset — live visits, visualized.
05Turbo C++, for old times' sake
Real C/C++ running in your browser — not a screenshot. Try F9.
Real execution via JSCPP, in a Web Worker — an educational C/C++ interpreter, not a full compiler: no STL containers, no classes, no goto.
06Nostalgic

Before Slack, before Discord — mIRC was how the world chatted in real time, one 56k modem at a time. Pull up a chair in #kaipaiton and say hi.
#kampung is a scripted hangout — the villagers are bots (BM only) and nothing leaves your browser. Switch to #kaipaiton for the live room. The real mIRC is still alive and updating, 30 years on.
07Let's talk
Got something to build or fix? Whether it's a site, a system, or a data mess — I can help.