Scott Rallya
Senior AI Systems Engineer | Retrieval & Reliability
New to Contra
Scott is building their profile!
When an AI workflow produces something that matters, “where did this come from?” should not be an afterthought.
I am developing a lightweight evidence-architecture approach for that problem: identify the sources, name the claims a workflow is allowed to make, show the review...
The AI workflow question I keep returning to is not “can it run?” but “can a person inspect what it is about to do?”
In Context Forge, I have been working on a small answer: make the task, sources, permissions, stop conditions, validation, and handoff visible before execution....
