I'm a Form 3 student in Sarawak, Malaysia. I like cybersecurity, local AI, and digital hardware. Most of my projects start because I want to understand how something works. Then I add tests and write down what the result does not prove.
- RecurQuant - a Qwen3.5 recurrent-state quantization research project with packed-state measurements and a frozen 500-task held-out fidelity result.
- ScamShield - a Flutter prototype made for the Young Innovators Challenge 2026 to explain Malaysian scam-risk signals.
- PortCVE - a Windows CLI for finding listening ports, their owners, exposure changes, and possible CVE matches.
- CyberRAG - a local cybersecurity retrieval experiment with MITRE ATT&CK grounding and a checked-in 15-question pilot.
- INT8 dot-product RTL - one small SystemVerilog compute core with a Python reference model and CI.
Some experiments worked. Some failed. I keep useful failure cases instead of turning every result into a success story.
I use coding agents substantially for implementation, debugging, tests, and documentation. I choose the problem, set the scope and claim limits, direct the iterations, review what gets published, and rerun the important checks.
I do not claim that every line was handwritten by me. If I cannot explain a result or show how it was checked, it should not be featured here.
Right now I am learning more about quantization, RAG evaluation, Windows/.NET, SystemVerilog, and how to design tests that are difficult to fool.
The LABEEB + Espeon contribution pattern is intentional pixel art, not a work log. Sprite source and attribution are in assets/README.md.