Gen.Urban: an in-car AI assistant, shipped into a running vehicle program
Three and a half months to absorb two years of multi-vendor architecture, build and integrate the TTS/STT AI player and generative AI into the driving platform, and hand it over well enough that it still powers the vehicle's demos years later.
- Client
- Prophet, for Volkswagen Group Innovation
- Year
- 2024 / Berlin
- Elapsed
- 3.5 months, concept to board-ready integration
- Role
- Built everything and directed: integration, orchestration, documentation, demo, maintenance handover
- Stack
- TTS/STT AI player in the existing HMI track / generative AI in the driving platform / multi-agent orchestration / py_mqtt_audio_player, now a licensed product
- Unlock
- Software still runs the vehicle's demos years later / re-established the VW relationship that produced the 2025 cockpit engagement
Full circle
Gen.Urban started in 2019 at VW Group Innovation's Future Center Asia, in the team I led from Beijing: HMI architecture, seating bucks, the first showcase generations. Five years later the program was heading for the road, out of time, and needing its AI layer built. The call came to me: I had architected the earlier generations, I was now on the outside, and I had brought VW in as a Prophet client.
The build
Three and a half months: I absorbed two years of accumulated multi-vendor architecture, built and integrated the TTS/STT AI player into the existing HMI track, wired in VW's ongoing AI efforts, built generative AI solutions into the driving platform, realised the personalised interior through multi-agent orchestration, and documented, demoed, and handed over maintenance.
Result
The board received it very well. The vehicle is on the road, and the software I built in those 3.5 months is still what runs when it is demoed, years later. And the relationship carried forward: in 2025 the same client came to me again, at a third firm, for the AI-augmented cockpit.