Security researcher & founding engineer
In my small personal lab, I research AI and security. As a founding engineer at VigilX, our side project, I turn that research into tools for investigating complex software.
I write about the decisions behind that work: what evidence a model missed, which assumption broke a system, and when changing the question led to a better design.
With the GovTech Edu team, I helped investigate a Google Cloud Armor flaw that led to a global patch. More recently, my research has found high-impact issues in heavily audited open-source systems.
VigilX · Founding engineer
The hard part of my AI research has become deciding where the next model call will help: supplying missing evidence, retrying an interrupted investigation, or challenging a surviving claim. The article follows the system behind those decisions, where it still falls short, and why findings need human review.
SHADOW · Data engineering
SHADOW began with a narrower question: which two lookups did the system actually need? That decision shaped a design for 4.19 billion canonical credential records—records of exposed login details—on a 16 GB Mac Mini. One warm, single-threaded domain lookup took 0.38 seconds. The article follows the design, the mistakes it exposed, and the limits that remain.
Google Cloud Armor · Security investigation
Cloud Armor’s logs made blocked and allowed requests look identical. After months of investigation, our team at GovTech Edu produced a reproducible report. Google confirmed the issue affecting Classic Application Load Balancers and released a global patch.
My October 2026 research summary covers 15+ reported security findings over six months, including high-impact issues in heavily audited open-source codebases.
What needs to stay predictable when the model itself isn’t.
Working out what to watch when visibility and budget are both limited.
Find my writing on Medium, projects on GitHub, or get in touch through LinkedIn.
