AI tools and workflow automation, from the first integration to the permissions, review steps, and operating docs your team needs.
Azure · Entra ID · Microsoft 365 · TypeScript · Python
Compare model outputs field by field, run validation checks, and send disagreements to a reviewer. Keep the original output alongside the final decision.
Extraction pipeline · Architecture and trade-offsA registry embedded in Teams and SharePoint, with role-based access, council review, and a record of each decision.
Governance platform · Case studyThe system I build to manage my own infrastructure. It brings service evidence, tasks, and reviewed actions into one operator interface.
Current architecture · September 2026Take an assistant looking after a document queue. Open each request to follow the decision. This is an illustrative policy design; it runs no tools and uses no live data.
queue.status for the assigned queue.Permission to observe a queue does not grant permission to change it.
queue.retry for one failed job.A timeout leaves the outcome unknown. Check the job before attempting another retry.
documents.export outside the caller’s assigned queue.An approval workflow cannot create access the reviewer is not authorized to grant.
I work directly with the team responsible for the system. We agree on the scope and deliverables before work starts.
Trace an authentication problem, examine a tool boundary, or work through a retrieval pipeline. Get a written set of findings and proposed changes.
Review packages and prices →Connect a model to business data and workflows, including permissions, failure handling, and the documentation needed to maintain it.
Discuss the scope →Ongoing design and code review for a team that wants another experienced engineer involved in the decisions.
Discuss ongoing support →I’m a solutions architect and software engineer in the Washington, D.C. area. My client work centers on Azure, Entra ID, Microsoft 365, and Power Platform, alongside TypeScript and Python.
I also build and operate my own systems. Spire is where I work through infrastructure problems that need sustained attention: service visibility, access control, and recovery when an action’s result is uncertain.
Send the current setup, the part that is stuck, and any deadline. A rough description is enough.