I created Silver State Cyber Solutions. SSCS is the name on the door, but it is me behind it, so this page is about what I actually do and what I can do for you.
I got my hands on a laptop around eleven years old and started taking apart a game I played. Other people had written add-ons for it and published the code behind them, and rewriting what they built is what actually taught me to program, because before you can change somebody else's code you have to read it and understand why it works. That habit never left me. Reading other people's code to figure out why it works is still most of what I do.
By fourteen I had moved to Linux and was running Kali, working through the tools in it and testing them in environments I built myself. Around the same time I got interested in the human side of the field, the questions about who holds access and what they are supposed to do with it. That is what decided my direction. I would rather be the person who closes the door than the person who finds it open.
My family told me for years to go get the degree. I put it off. When I finally enrolled it went differently than I expected. The coursework did not slow me down, it gave me structure, and I started building outside of class faster than I ever had on my own.
Purdue Global University, Bachelor of Science in Cybersecurity, graduating January 2028.
Most of my proof is public, so you can go verify it instead of taking my word for it.
I built an isolated VirtualBox environment for running untrusted AI agent code, scripted end to end in PowerShell so it rebuilds identically every time. The verifier asserts that network egress is genuinely cut and diffs the machine's full configuration key set against a committed baseline. Both checks are negative tested, meaning I proved they fail when they are supposed to fail. A check you have never seen fail is not a check.
The ai-security-lab repository on GitHubI contribute to NVIDIA's NeMo agent framework. I filed an issue on out-of-sync test documentation that a maintainer picked up and resolved. My most recent merged pull request fixed a bug in their tracing test suite where a mocked HTTP response sent the test down the failure path instead of the success path its name described, so it passed without ever testing the behavior it claimed to cover. The fix took that file's runtime from about five seconds to under a tenth of a second.
NVIDIA's NeMo agent framework repository on GitHubI wrote up the false pass I found in my own verifier, then wrote a follow-up when OpenAI and Anthropic both disclosed the same failure at production scale.
I am also working through the OWASP Top 10 for LLM Applications and MITRE ATLAS. Agent security is where I think this field is going and I would rather be early than caught up.
I do three things. An external exposure review finds what an attacker can already see. Fixing what the review finds closes it. Ongoing checks keep it closed.
The hardest part of this work is not technical. Most people do not know what they are exposed to, so they do not know there is anything to fix. So I will tell you plainly what I found, what someone could do with it, and what it would take to fix. You decide from there.
I bartend four nights a week to pay for school, and I am not going to hide that. Working a high volume bar teaches you to read people quickly and stay steady when the room is not. That is the same muscle social engineering runs on, and it is part of why I take the human side of security as seriously as the technical side.
Email me at yeriahz@sscsnv.com. Everything above is linked below so you can check it yourself.
Email me and we will set up a free consultation. Tell me what you run and what you are worried about, and I will tell you whether there is anything worth doing.
yeriahz@sscsnv.com