Taxi apps / ridesharing

The original platform had outgrown its design: dispatch, bookings and the partner programme had each been extended past what the architecture assumed, and the whole thing ran on hardware in a room someone had to visit. Human dispatchers spent their shift on the same three questions — ride requests, cancellations and 'where is my car' — leaving no time for the calls that actually needed judgement.

Rework the stack and the main operational flows together before putting voice on top. Migrating the same complexity to the cloud would have bought nothing, so the component count came down as part of the move. An AI dispatcher is only as good as the data it can read and the dispatch it can trigger, so the services came before the conversation.

Backend on C# / .NET and Symfony, infrastructure in the cloud, WebSocket communication with vehicles and driver apps for live location, ride state and dispatch events. Availability now sits at 99.9% and API response times below 100 ms, on a platform the operations team can reason about.

A voice agent wired into telephony handles requests, cancellations and information end to end — handing over to a person when it should — while unified WebSocket communication keeps vehicle, driver, dispatch and customer applications on one channel.

C# .NET and Symfony services, Node.js edges, PostgreSQL and MySQL, Redis, WebSockets, cloud infrastructure, telephony integration, ElevenLabs-style voice orchestration, and CI/CD pipelines that deploy on merge.