AI Product Proof
A local demo makes sources, conflicts and missing evidence visible.
Local synthetic reference demoLocal synthetic reference demo
The local demo uses only approved synthetic sources.
Sources and their limits are made visible in the report.
A supported answer, a blocked claim or no answer.
Human reviewA person reviews the result. The demo does not publish or perform external actions.
Local reference demo with synthetic data; no generative model integration.
Problem and result
Document based AI workflows can sound convincing even when a source is outdated, a capability is only planned or evidence is insufficient.
Build a local control and evaluation layer before adding generative model variability.
Implementation was accelerated with AI coding agents. Product framing, scope, acceptance criteria, review and release decisions remained my responsibility.
It is not a production assistant and does not include a generative model integration.
Examine the work
The workflow diagram and this case description explain the current state. Source code and technical documents are in a private repository; there is no publicly runnable demo here.
Current status
Interview ready local reference demo with synthetic data and validated product boundaries.
Every final decision remains review required. The system can classify and block claims, but it does not publish or act autonomously.
A reviewer needs to know whether an AI workflow can separate implemented release facts from planned capabilities, conflicting evidence and missing evidence.
Hiring managers, AI Product Leads, data platform Product Managers, Engineering Leads and technical evaluators.
The milestone isolates retrieval, answer context, claim classification, uncertainty, blocked claims, static validation and human review.
Implementation was accelerated with AI coding agents. Product framing, scope, acceptance criteria, review and release decisions remained my responsibility.
Every final decision remains review required. The system can classify and block claims, but it does not publish or act autonomously.
Interview ready local reference demo with synthetic data and validated product boundaries.
A real reviewer or product team should use the demo in an interview, design review or controlled evaluation and record what was clear, unclear or missing.
These commands apply to the private source repository, not a publicly runnable web demo. With authorized repository access, the local demo runs without network, database or LLM access.
npm run -s jarvis:reference-demo
npm run -s jarvis:reference-demo -- --interview
npm run -s jarvis:reference-demo -- --json
npm run -s jarvis:reference-demo:validateWebsite care & technical security maintenance
For your new or existing website: a focused review, agreed improvements and a handover you can work with. Ongoing care can be arranged to suit the site.
Review known security advisories, test suitable updates and check important visitor journeys. Receive a clear account of completed work and remaining issues.
Keep forms focused, load external content deliberately and match consent to the feature. Data flows and required provider agreements are clarified before use.
Access, responsibilities and recovery belong in the handover. Ongoing care has an agreed system scope, schedule, budget and response times, so you know who handles what.
The deliverable is documented technical work. Legal review, independent penetration tests and 24/7 incident coverage are separate services. Privacy decisions and approvals are agreed with the responsible operator.
The next step
Describe the task, the people involved and the outcome you need. A few sentences are enough for an initial conversation.
A non-binding enquiry. We agree scope, price and acceptance before an assignment.
Please leave confidential documents out of your first message.
Or by emailhello@wateristheholygrail.com