A production connected-vehicle platform, architected by me and built with AI

A client project that grew into a multi-tenant platform for e-mobility manufacturers and fleet operators. It is live, managing more than 12,000 vehicles, and it shows what an experienced architect can now deliver with AI tools without skipping security testing, recovery, data protection or regulation.

The brief

A client launching a new e-mobility business needed a connected-vehicle platform: something to follow each scooter from the factory to the end of its life, support the people who sell, repair and ride it, and satisfy the regulators in every market it would be sold into. It had to work from day one, with no existing systems to lean on.

What was built

The platform covers nine stages of a vehicle's life: factory quality checks, distribution, registration, active use, diagnostics, anti-theft, service and firmware, transfer or resale, and end of life. It has four surfaces, each branded for the customer:

• An operator portal for customer service, fleet operations and administration: a fleet map, fault-code lookup, battery-health scoring, telemetry analysis, notifications, and supply-chain tracking from factory to customer.
• A workshop portal, designed for tablets, for job management, diagnostics, scooter set-up and firmware management.
• A customer app for iOS and Android, with Bluetooth pairing, live speed, battery and range, ride history, sharing, remote locking and theft alerts.
• A customer web portal for the same account in a browser.

Some of the tools come straight from the engineering work I advise on. The range simulator uses Monte Carlo modelling to show the realistic range a rider can expect, rather than a single optimistic figure. Battery-health scoring flags failing cells before they turn into returns.

From one client to many

What began as one client's platform became a multi-tenant product that e-mobility manufacturers and fleet operators can run under their own brand. Each organisation has its own instance and its own data: per-customer databases keep data in the region and jurisdiction it needs, and per-region feature flags turn capabilities on or off by country across 18 regulatory zones, including age limits. Approved regional users can edit their local wording themselves, so the platform can launch in a new market without a software release.

Security, safety and data protection

• Four rounds of penetration testing and a dedicated Bluetooth security review.
• 4,000 automated security tests across the database, web, APIs, apps and Bluetooth connections, run with industry-standard security tools by different AI models from the ones that wrote the code.
• Signed over-the-air firmware updates with SHA-256 validation and rollback, so a failed update cannot leave a vehicle unusable.
• Location recorded at city or town level only; precise GPS positions are not collected.
• GDPR built in: consent by category, withdrawal tracking, data export, and a seven-day grace period before deletion.
• Reviewed under a load of 300,000 simulated users.

How it was built

Fienti was heavily AI-enabled and closely guided by me as architect. AI coding tools did much of the typing; the engineering decisions were mine: the multi-tenant and multi-region design, the security model, how firmware updates fail safely, and what to test and how. It builds on the multi-region pattern I developed earlier for a global e-mobility business.

I wrote about the experience in When AI removes the translation layer and The hobby system and the professional system, and about the security side in Connected products versus AI-assisted attackers.

Where this helps you

If you are adding connectivity to a product, reviewing a connected platform before you back it, or weighing up whether to build or buy, this is the kind of work I can help with. See connected products and independent advice and due diligence, or book a 30-minute conversation.