Alexandria Hamilton spent seven years inside the electronic health record industry, the software that holds everything a hospital knows about you. That was long enough to watch artificial intelligence move into clinical work, and long enough to notice that almost nothing was checking its answers.
So she built the thing that checks. Orinyx, described in a report published September 29, 2026, is what she calls an independent safety layer for clinical AI. When a hospital’s system produces a result, Orinyx holds that output against FDA labeling and clinician guidelines and looks for what does not match: a hallucination, a claim the label does not support, information that is simply wrong. It runs outside the system it is checking, and that is the whole design. A model grading its own work is not oversight.
Hamilton describes herself as an AI systems architect and founder. She built Orinyx under charmthirteen, her own AI advisory and systems design practice. The product is in beta.
There is no funding round in this story. No Series A, no valuation, no investor list to recite. What there is instead is a woman who knew an industry from the inside and decided the gap in it was hers to close. She is currently taking a pathology class.
“I really just want to deepen my clinician knowledge, continue to build things, and just build AI products that are for good,” she said.
That last part carries weight in a field where the marketing has run well ahead of the evidence. Hospitals are buying clinical AI now. The tools learn from records that already hold decades of decisions about who gets believed, who gets tested and who gets sent home, and Black women have documented reason to be careful about what a clinical system concludes about them. An AI trained on that record does not correct it. It repeats it faster, and with more confidence in its voice.
Checking the output against the label and the guideline is one of the few places that can be caught before it reaches a person. Hamilton is building that check in beta, out of her own practice, while sitting in a pathology class to better understand the clinicians she is building for.
That is what the work looks like before anybody writes a check for it.
