Configuration, installation, entitlements, and local runtime requirements.
Platform build
RegulatoryModels Platform
The implementation track behind the regulated AI sandbox: control planes, client environments, submissions, sandbox execution, evidence packages, and reviewer workflows.
01 / Problem
A sandbox concept needs platform foundations before regulated users can trust it.
RegulatoryModels moves from product idea to implementation details: how clients are configured, how data and environments are isolated, how submissions are represented, how sandbox jobs run, and how reviewers receive evidence they can inspect.
Datasets, environments, RBAC, submissions, and audit records.
Sandbox jobs, sample models, runs, and results.
Decision records, evidence packages, lineage, and reviewer exports.
02 / Architecture
A phased platform architecture for accountable model evaluation.
The repo is organized around phased plans: foundation architecture, phase-1 control plane, phase-2 sandbox MVP, phase-5 evidence and decision records, and phase-6 pilot readiness. It includes frontend, backend, scripts, sample data, sample models, and implementation reviews.
Use case brief, submission contract, environment design, qualification policy, and system architecture.
App shell, auth, RBAC, entitlements, schema, seeds, and dataset environment flow.
Sandbox runner, backend API, sample model, sample data, and prototype walkthrough.
Evidence package schema, reviewer export, demo guide, QA, and design-partner handoff.
03 / Data
The repo defines client, environment, dataset, submission, run, result, entitlement, and evidence data surfaces.
The docs include schema and seed specs, data model documents, sample data specs, environment configuration guides, evidence schemas, and lineage maps.
04 / Prototype
The platform has a local app and implementation documentation.
The root `index.html` and app shell can be opened locally. The repo also includes backend docs, sample data, sample models, and scripts for the sandbox runner path.
Open local RegulatoryModels app05 / Learned
Accountable AI products become real through boring infrastructure.
The implementation docs make the sandbox idea more concrete. Trust depends on schemas, entitlements, installation paths, audit maps, exports, and QA notes, not just a polished evaluation UI.
- The sandbox concept needs a control plane to be deployable.
- Evidence and reviewer workflows should be first-class platform objects.
- The implementation track gives the product story technical credibility.