Engineering showcase
Try the engineering, not just read about it
Working prototypes and concept labs that run in your browser with fictional data. Each one says what is implemented, what is simulated, and what connecting it to a real business would require.
These are standalone public demonstrations of concepts, separate from Donna’s confidential client prototypes. None of them reproduces a client’s system or a private production system, and none proves the outcomes of one. Real integrations connected: none.
Working prototypes
-
AIUNIQ automation · prototype
Workflow Automation Builder
Configure required fields, routing rules and approval for a fictional enquiry, run it, and download the resulting work item as JSON and as a readable summary.
- Implemented
- Workflow checks, validation, ordered routing with reasons, approval, history, work item and downloads
- Simulated
- Email, CRM, booking and queue actions
-
Content workflow · prototype
Content Workflow Prototype
Turn a fictional brief into a plan, check a draft's claims against its numbered sources, request revisions and see the gate decision: ready, needs human review, or blocked.
- Implemented
- Brief checks, the plan, deterministic claim checks, capped revisions, the gate and the reviewer's decision
- Not connected
- A writing model: sample drafts are hand-written, template drafts only quote your sources
- Related concept
- Evidence-checked content with a named outcome, as in the governed-content case study and AIUNIQ ContentOS's plans. It uses no client material and reproduces neither system
-
Private research prototype · recorded results
RAG Evaluation Explorer
Compare retrieval providers on retrieval, answers, citations and abstention across seven recorded evaluations under three pre-registered protocols, and step through invented teaching examples. No evaluated provider met every quality target.
- Recorded
- Run-level measurements, targets and met or missed marks, copied from archived reports and checked against their hashes
- Illustrative
- The teaching examples, invented for the page; nothing is retrieved or answered live
-
AIUNIQ automation · walkthrough
Enquiry Automation Walkthrough
Six fixed fictional enquiries through intake, validation, a routing decision, human approval where required and a simulated result. A browser simulation, separate from the AIUNIQ Enquiry Desk application.
- Implemented
- The step engine, policy, approval bound to the plan, redaction
- Simulated
- Email, notifications and CRM updates
Concept labs and the projects they relate to
Each lab demonstrates a concept from a project on the portfolio. It is a standalone reconstruction with fictional data: it does not reproduce that project's system, and it does not prove that project's results.
-
Lab 01 · concept lab
Privacy Boundary Lab
Identifiers become stable tokens before text would reach an AI service, and are restored only after authorisation.
- Related project
- Privacy & Intelligence Boundary, a submitted technical assessment (design, not a production deployment)
- Simulated
- The external AI service
-
Lab 02 · concept lab
Governed AI Run Simulator
A fictional brief moves to a publishing decision: ready, needs human review, or blocked.
- Related project
- Governed Legal Content Command Center, sanitized client work; the lab uses none of its stages, data, prompts or metrics
- Simulated
- The AI stages and reviewers
-
Lab 03 · concept lab
Independent Review & Evidence Ledger
Simulated reviewers raise evidence-linked findings, and a person makes the final call before a gated release.
- Related project
- Governed Legal Content Command Center (the independent-review concept only)
- Simulated
- The reviewers
-
Lab 04 · concept lab
Reliability and Recovery Lab
Failures, bounded retries, duplicate protection and recovery from a checkpoint without repeating finished work.
- Related project
- The reliability concerns in Donna's governed-workflow work; no specific system is reproduced
- Simulated
- The delivery service and its failures
-
Lab 05 · concept lab
AIUNIQ Control-Plane Demo
Jobs admitted against a budget of fictional units, approval of the exact revision, and operator pause, resume and cancel.
- Related project
- A standalone AIUNIQ demonstration. It is not a reconstruction of AIUNIQ ContentOS, whose Phase 0 has no approvals or budgets
- Simulated
- Jobs, budget units and operators
-
Lab 06 · concept lab
Fleet Operations Lab
A simulated fleet of 5 to 20 agents works through dependent tasks, bounded retries, checkpoint recovery, a budget ceiling and a human approval checkpoint.
- Related project
- Multi-Agent Fleet Orchestration on OpenClaw, Donna's design-stage AIUNIQ architecture. Nothing of it is built on OpenClaw yet; OpenClaw is a third-party open-source project, and no affiliation is implied
- Simulated
- The agents, tasks, faults, tokens and costs; OpenClaw is not connected
Voice guide
Local preview only · not on this site
AIUNIQ portfolio guide, spoken
The same approved answers as the text guide, read aloud, with a visible transcript, a stop control, session limits and a text fallback. The speech provider keys stay on a server, never in the page.
- Status
- Runs only on Donna's machine. Tested with a fake speech provider and a fake microphone; not yet tested with a real speech service.
- Before visitors can use it
- A licensing decision for the speech provider, approved usage limits and costs, a hosted server, a privacy notice and a reviewed security policy for that page
Until then, use the text guide: open the portfolio and choose “Ask the guide”.