Donna G · AI Systems Engineer
I build AI systems that earn the right to act.
I build AI systems and automations you can check before anything important happens.
- Automate repetitive work such as enquiries, documents and hand-offs, with a person approving before anything is sent.
- AI assistants that stick to approved facts and say when they don’t know.
- Keep sensitive data protected before it reaches an outside AI service.
Working demonstrations
Try it in your browser
Six short demonstrations, all with fictional data and simulated or local logic. Nothing is sent anywhere, and none of them is a production system.
-
Privacy Boundary Lab
Names and emails are swapped for safe tokens before text would reach an AI service, then restored on your side.
Local deterministic demo Open lab: Privacy Boundary Lab -
Governed AI Run Simulator
Follow a fictional brief to a publishing decision: ready, needs human review, or blocked.
Interactive reconstruction Open lab: Governed AI Run Simulator -
Independent Review & Evidence Ledger
Simulated reviewers raise findings against evidence, and a person makes the final call.
Fictional simulation Open lab: Independent Review & Evidence Ledger -
Reliability and Recovery Lab
Failures, limited retries and recovery from a checkpoint, without doing the same work twice.
Fictional simulation Open lab: Reliability and Recovery Lab -
AIUNIQ Control-Plane Demo
Jobs admitted against a budget, approval gates and operator controls.
Fictional simulation Open lab: AIUNIQ Control-Plane Demo -
Enquiry Automation Walkthrough
A fictional enquiry goes from intake to human approval to a simulated result.
Simulated actions Open walkthrough: Enquiry Automation Walkthrough
My operating principle
Models can propose.
Systems must decide.
My work sits where probabilistic intelligence meets deterministic control.
I design orchestration, provenance, review, privacy, and publish boundaries that make AI useful in consequential workflows. I am strongest when a product needs both speed and a clear answer to: “What happens when the model is wrong?”
Proof, not adjectives
Systems I have shaped
Real projects, described honestly: what was built and verified, what was design work, and what is still planned. Open any case study for the full technical detail.
Think in flows
Explore the architecture
Choose a system, then inspect each boundary to see what it owns and why it exists.
Choose a stage
Click any node in the architecture to inspect its responsibility and control boundary.
Interactive proof · preview
See a privacy boundary at work
A short preview of the Privacy Boundary Lab: fictional names and emails become safe tokens before text would reach an AI service, then are restored on your side. Same engine as the full lab.
Open the full Privacy Boundary Lab All interactive labs Enquiry automation walkthrough
Deterministic tokenization
Demo authorization is assumed in this preview; this is not real authentication.
Not yet protected.
Protected payload will appear here.
Selected timeline
From support to systems
A career shaped by operating real systems, diagnosing failure, and building better boundaries.
How I contribute
Architecture through acceptance
Hiring, or planning an AIUNIQ project?
Need an AI system—not just an AI feature?
I’m open to AI engineering, orchestration, governance, privacy engineering and technical integration roles, and I take AIUNIQ enquiries for AI engineering and workflow automation projects. Email is the quickest way to reach me.
