Ride sharing across two markets

Context

I have built ride sharing in two markets that could not be more different. HipCar in Moldova: a young market, where the problem is trust — verifying that the vehicle exists, the documents are valid and the driver is licensed. Waave in the US: a mature, regulated market, where the problem is compliance — a phone acting as a legal taxi meter has to be certified like a fuel pump.

Same category, opposite constraints. That contrast is most of what I know about ride sharing that a slide deck cannot teach.

Constraint

In Moldova the constraint was verification without friction. Onboarding had to confirm real drivers and real vehicles against official sources, same-day, at signup volumes that rule out manual review — in a market where every hour of delay loses the driver to the alternative of just not joining.

In the US the constraint was regulation with no negotiation. The National Type Evaluation Program certifies measuring instruments, and a taximeter app either meets the accuracy bar or it cannot legally charge a fare. There is no growth-hacking a weights-and-measures authority.

Decision

At HipCar we rejected manual document review and integrated external verification services directly — vehicle existence, document validity, driver licensing, checked against the source rather than a photographed PDF. Slower to build, permanently faster to operate.

At Waave we rejected the certified hardware meter in every car and certified the app itself instead. Hardware is the easy certificate and a permanent per-vehicle cost; software certification was the slow, uncertain path that removes a hardware line from the unit economics forever. Measurement error came in under 0.1%.

Across both, we rejected assuming that what worked in one market transfers to the other. The identity stack built for Moldova would not satisfy a US regulator, and US compliance overhead would have strangled Moldovan onboarding. The shared layer is the dispatch, pricing and telemetry logic; the trust layer is local.

What shipped

HipCar in Moldova: ride sharing with driver KYC wired into external registries, alongside the Citylink car-sharing operation. Waave in the US: a ride-hailing platform with an NTEP-certified software taximeter, a rebuilt AWS foundation for investor due diligence, and an AI roadmap covering demand prediction, routing and dispatch agents.

Numbers

Two live ride-sharing markets. NTEP certification with measurement error under 0.1% at Waave. Six years as Waave CTO, 2019 to 2025, across New York and Boston. Seven years across HipCar and Citylink, 2018 to 2025.

What I'd do differently

I would have written down, per market, the list of things that must not be shared between codebases. We learned the boundary between portable platform logic and local trust logic by crossing it a few times, and each crossing was a migration.

At Waave I would have started the certification test harness before the meter logic itself. Tuning accuracy against a measuring setup that is itself still changing makes a slow process slower — and certification is never fast.