Apply
Research runs on provenance. Yours should too.
Your methods section says what you did. The record of it is scattered across chat histories, downloads, prompts that are gone, and a folder named final-v7. When the committee, the reviewer, or a co-author asks exactly what was searched and who decided, the answer should not depend on anyone’s memory.
Why
I built this for enterprises. My doctorate needed it more.
OrbisFramework was built to record how an AI decision gets made: the workflow, the gates, the human judgment, and the trail. Enterprises need that because a board or a regulator eventually asks how the system decided. Then, in 2024, I started a doctorate, and found the same question waiting there, asked by a harder audience. A committee does not ask whether you used AI. It asks how you know what you claim to know.
A research project run with AI generates hundreds of judgments: what was searched, what was screened out, what was rewritten, who decided. Keep them in folders and spreadsheets and you will reconstruct your own project from memory in front of the people examining it. Keep them on a ledger and the answer to how do you know is: here, timestamped, attributed, in the same place my advisor and co-authors work.
So the infrastructure under my own doctorate is the infrastructure on this page. No output advances without my judgment, the models underneath are routed and replaced as better ones arrive, and the record accumulates. The argument stays mine. That is the whole design.
What it does
Every action is on one ledger. Timestamped and attributed, in the same place your advisor, your committee, and your co-authors work. No spreadsheets, and no reconstruction.
Automation with a human deciding. No AI output advances without the researcher’s judgment, so the argument stays the researcher’s. That is what makes it defensible in a committee room and publishable after.
Models are routed and replaced underneath. The workflows, evaluations, and records accumulate while the models underneath change, so the lab keeps what it learns instead of starting over with every release.
It is the infrastructure OrbisDiscovery runs on. The 32 step lifecycle executes on this foundation. The two pages describe one system, not two products.
The question is never whether you used AI. It is whether you can show how.
Thirty minutes. Bring the project, and leave knowing what its ledger would look like.