Work / Optimizing, integrating and deploying AI models on automotive embedded hardware.AI deployment for automated driving
Context
CARIAD is the Volkswagen Group software company, in Berlin. Since April 2023 I have worked there full-time as part of the Automated Driving Alliance between CARIAD and Bosch, building an ADAS software platform for Volkswagen Group vehicles on Qualcomm SoCs.
The problem
A model that performs well on a workstation still has to fit a fixed memory and power budget on automotive hardware, and meet a latency target every frame. Deployment also has to meet ISO 26262 functional safety and ASPICE process requirements, with reproducible results across several platforms.
What I did
As Head of AI Deployment (October 2025 – present) I lead a team of 12 engineers that takes AI models from research prototypes to production on the volume-production ADAS platform:
- Optimize, integrate and deploy AI models on embedded automotive hardware.
- Own production reliability, performance and auditability: latency, memory and power targets, release quality and traceability, under ISO 26262 and ASPICE.
- Work directly with enterprise stakeholders at Volkswagen Group brands (Volkswagen, Audi) on requirements, priorities and delivery.
- Lead technical discussions with semiconductor partners Qualcomm, Samsung and NVIDIA on integration and performance.
- Drive shared tooling and automation so model teams deliver consistently across multiple platforms.
Earlier, as Senior ML Software Engineer, I was Technical Product Owner for the AI Runtime Environment. I built a Python SDK for model conversion, quantization and optimization, plus a runtime abstraction layer. I also integrated and deployed prediction, occupancy and sensor-fusion models on Qualcomm AI accelerators.
Stack & notes
Python · Qualcomm NPU · ISO 26262 · ASPICE · embedded systems · automotive standards.