Offensive Security · Artificial Intelligence
Thirty years in technology, more than twenty of them spent breaking things on purpose — now working where that expertise meets AI.
Footage eberhard grossgasteiger / Pexels
I'm a cybersecurity professional, author, instructor, and builder. My career started in systems and infrastructure — Windows Server administration, large enterprise environments — before moving deeply into penetration testing, red teaming, adversary simulation, and security engineering.
For years my job has been to test systems the way real attackers do: finding weaknesses, developing attack paths, and helping organizations understand and reduce their actual risk rather than their theoretical risk.
Teaching has always been part of how I approach this work. I've authored cybersecurity certification material, including books for CISSP candidates, and spent years as a technical trainer, consultant, and mentor. I like taking complex subjects apart and finding better ways to explain not just how something works, but why it works.
Much of my recent work sits at the intersection of artificial intelligence and security. I run substantial local AI infrastructure and experiment constantly with local LLMs, coding agents, and orchestration systems — trying to find where AI genuinely improves technical work, and where it only appears to.
I'm less interested in replacing expertise with AI than in what experienced practitioners can do with it — how much more they can research, build, test, and learn.
An evolving cybersecurity platform and laboratory — hands-on offensive-security challenges, training, and experimentation, built around the growing intersection between security and artificial intelligence. It's where most of my ideas get tested against something real.
Explore AtomicCTF →A kernel-mode denial-of-service affecting a widely deployed commercial Windows endpoint-detection-and-response agent. Triggerable by an unprivileged local user, taking the host down and the EDR offline with it.
Reproduced in a lab and disclosed to the vendor, with a follow-up report covering their latest release.
Check Point VPN-1 SecuRemote/SecureClient NGX cached credentials under a registry key readable by Everyone, allowing a local user to recover and reuse them.
Reported and credited to Mike Vasquez in 2008.
CVE record →58th of 686 teams while competing solo.
Scoreboard →Additional research and technical write-ups will be published as disclosure timelines permit.
A learning and assessment system built around a simple idea: teach people what they need to learn instead of making everyone sit through what they already know.
Traditional training often assumes every learner starts from the same place, while conventional testing can measure recognition and test-taking as much as genuine understanding.
This system uses authoritative source material to assess what a learner actually understands, identify individual knowledge gaps, and create targeted learning and practical work around what they need next. Learners can then be reassessed as their knowledge develops.
The initial development target is Python and technical learning, but the underlying problem exists anywhere organizations need people to genuinely understand what they're being taught.
In active developmentLocal LLMs, multi-agent orchestration, coding agents, model routing, and security experimentation — substantial local AI infrastructure used to find where these systems genuinely help and where they quietly fall apart.
I approach technology as both an engineer and an experimenter. If something interests me, I want to build it, break it, automate it, measure it, and find out where its limits are.
This is where I document that work — projects I'm building, security and AI research, technical experiments, lessons learned, and occasionally things I build simply because I want to know whether they can be done.
Security research, AI security, technical training, consulting, and interesting collaborations.