Architecture & AI governance

The model can reason.
The architecture decides.

I design the boundaries around AI: who can approve, what evidence counts, when permissions expire, and how the system stays truthful when something fails.

The questions that shape the design

Judgment before a stack.

I treat patterns as hypotheses, test them against evidence, and connect the architecture to the decision the business needs to make.

01Where is the business actually stuck?

Start with the workflow, the time spent, the bottleneck, and the consequence of a wrong decision. Identify where AI changes the outcome and where a simpler control solves the problem.

02Who should have authority?

Separate recommendation, approval, identity, and execution. Bind approval to the exact decision, scope the permission, and keep the model from granting itself authority.

03What evidence has to be true now?

Check fresh external state and current policy at the execution boundary. Retain provenance and exact artifact identity so yesterday’s approval cannot silently authorize today’s different action.

04What happens when the system is wrong?

Design rejection, retry, recovery, compensation, and observability alongside the happy path. Preserve failed attempts. State should follow verified evidence, and recovery should preserve the history of the original decision.

One boundary, four responsibilities

Make control visible in the system.

These responsibilities appear across the selected projects. Each implementation states what it actually proves.

01 / PROPOSE

Reason over the problem.

The AI produces a constrained candidate with supporting evidence.

AI capability

02 / AUTHORIZE

Bind the decision.

A person approves the exact terms within an explicit scope.

Human authority

03 / VERIFY

Check current reality.

Deterministic controls check policy, evidence, and permission at execution.

System control

04 / RECORD

Keep the result truthful.

Execute or reject, preserve evidence, and support review and recovery.

Business accountability

Inspect the authority evidence

Working with me

Close enough to build.
Senior enough to shape the direction.

My background spans financial operations, cloud platforms, and hands-on AI engineering. I lead work from discovery and architecture through delivery, with the business consequence kept in view.

tmanning@post.harvard.edu

Connect on LinkedIn

Professional focus: AI Architect · Cloud / AI Architect · Enterprise AI platforms · AI transformation consulting
Location & mobility: Based in Austin. Willing to relocate within Texas, including the Texas Triangle. Open to meaningful business travel.
Work style: In-office or hybrid; direct collaboration with engineering teams and executive stakeholders.
Current practice: Founder & AI Architect, Apex AI|ML Engineering LLC. Enterprise engagement details remain confidential.
Technical foundation: Python · AWS / GCP · SQL · Terraform · IAM / OIDC · Docker · CI/CD · Governed agentic systems