FruggrFruggr

An organisation inventories its AI systems, assesses their EU AI Act compliance and its own AI maturity, runs an action plan, and measures its real use of AI. I laid the foundations of the production application and led its build; it is live with its first paying customers.

• The production foundations: a typed API in NestJS and tRPC, authentication shared with the rest of the suite, and every piece of data isolated per organisation.
• The first business screens: the AI system registry, the maturity assessments and the action plan.
• The MCP server, with OAuth 2.1, so AI assistants can work with the platform.
• What makes AI usage telemetry useful: rules that tie each use to its AI system with a live preview, and production telemetry replayed on test environments so every screen holds the volume.
• An access policy required on every entry point of the API, with tests that prove the isolation between organisations.

The module audits websites against the French accessibility standard, RGAA, from a Chrome extension or a scanner, by hand or with the help of AI.

• The accessibility scanner, taken to production in a container: audits behind a login, hardened images, and a client token that never leaves for the audited site.
• The automatic mode of the AI-assisted audit in the extension, which runs a whole audit on its own, and the extension's releases.

The module analyses the footprint of websites and mobile apps, and their compliance with ecodesign standards.

• The ecodesign compliance forms, the French RGESN among them, with an engine that pre-fills criteria from each analysis.

• Personal access tokens and the authentication chain that let clients' CI/CD call Fruggr, and silent single sign-on over SAML.
• Administrator of the product's Azure infrastructure, in charge of its DevOps work through to production.
• Co-author of the project's technical constitution, followed by the team and by AI agents alike, including the rule that no agent merges without a human's approval.
• Rotating customer support duty, with investigations started by an AI connected to the product's data.