An intelligence layer for electric fleets

01/The challenge

A growing electric two-wheeler fleet produces an enormous amount of signal — battery state, location, charge cycles, rider behaviour, fault codes — and almost none of it is useful in raw form. Operations were being run reactively: a vehicle failed, someone noticed, a technician was dispatched. At a few dozen vehicles that is survivable. At a thousand it stops working entirely, and every hour a bike sits idle is rental revenue that never arrives.

02/The solution

We built an intelligence layer that sits above the hardware. Telemetry streams off every vehicle into a time-series pipeline, where it is normalised, scored and turned into decisions rather than dashboards. Fleet managers get live state across the whole estate; the rental flow handles allocation, handover and return without paperwork; and a maintenance model watches degradation patterns to flag a battery or drivetrain before it strands a rider. Alongside it we shipped a rider app for unlocking, live range, trip history, charging locations and support — the fleet's public face, and the thing that decides whether a rider comes back.

03/The results

Operations moved from reactive to scheduled. Faults are caught while the vehicle is still on the road rather than after it stops, utilisation is visible per vehicle instead of estimated in aggregate, and the team can add vehicles without adding proportional headcount to watch them.