I spent 25 years making machines trustworthy. Now they run themselves.

For most of my career, the systems I built did what they were told. You wrote the rules, the software followed them, and when something broke you could trace it to a line you'd written. Predictable, and that predictability is what made it safe to hand a real business with real money and real customers.

That world is gone. The AI agents leaders are adopting now act on their own and learn on their own. They make calls nobody programmed. So here's the new problem, and it's the one NourFide exists to answer: when an autonomous system decides for itself, how do you know it decided well, and how do you prove it to a customer, a regulator, or yourself? That question, AI oversight you can actually trust and verify, is what I write about.

The builder's lens

I'm not writing about AI governance from the outside. I've spent 25 years making systems trustworthy in production, which is harder than making them work. Builder and operator, not commentator.

My first job out of school, in 2000, was validating chips at PMC-Sierra. The test was brutal by design: sweep voltage, temperature, and frequency across their whole range and find where the chip failed. And one clean run proved nothing, it might be an outlier, so you ran it about thirty times across multiple chips before you trusted the result. By hand, that was months. I automated the rig, every instrument synced, one click, forty-five seconds a run. I didn't make the testing less rigorous. I made the rigor cheap.

After that I ran my own business building custom automation for other companies, the same job every time: a process that's manual, slow, or breaking, and the system that fixes it. Online learning platforms, a custom social platform, intake and back-office automation, across nonprofits, law practices, startups, businesses of every size. Different problems, one discipline.

The thread: I made automated systems trustworthy before "AI governance" was a phrase anyone used. The principle that kept a chip honest in 2000, don't trust a result until you've proven it across the full range, is the same one that governs autonomous AI agents now. Most people writing about it have never had to answer for a system that failed in front of a paying customer. I have.

---

I've written a book on this exact problem: how you build oversight for AI agents that keep learning after you deploy them, oversight the system itself can't quietly subvert. The manuscript's finished. It's the argument I'd been circling for years, finally written down.

What I'm working on

The foundation is a B.S. in Computer Science from Penn State, 1999, and 25 years of building since. The rest is keeping the edge sharp: I'm finishing a Master of Science in Artificial Intelligence at Woolf University, delivered through the Udacity Institute of AI and Technology. I completed UT Austin McCombs' Post Graduate Program in AI and Machine Learning (August 2026), and I'm working through a second in Generative AI now.

I also hold patent-pending research in AI governance. The specifics stay under wraps while the patent is pending.

Why "NourFide"

The name joins two ideas. Nour is Arabic for light. It's also my daughter's name. Fides is Latin for trust. Put together: light you can trust.

Not light that dazzles, not light that sells. Light you can rely on, that shows you what's actually happening inside the AI systems most people are told to take on faith. I don't think leaders should take them on faith. I think they should be able to verify. NourFide is my argument for how.