
Vision ASM
Open-source AI blind-spot monitoring for StarPilot vehicles.



A safety-critical, community-driven vision system that detects vehicles in adjacent lanes using the driver-facing wide-angle camera. It ships in StarPilot's stable release and runs in real time on embedded automotive hardware.
- Appointed to Development Team with full repository write access to help steer project roadmap and backend architecture.
- Authored and merged 1.7K+ lines of C++/Python into production, deploying a real-time blind-spot monitor to thousands of active vehicles.
- Trained custom YOLO26n ONNX models on 27+ hours of crowdsourced driving video with semi-automated CAN-bus radar labeling.
- Engineered edge performance on Snapdragon SoCs using zero-copy NV12 buffer slicing and CPU core isolation under 1.2x throttle overhead.
- Drove community adoption (27K+ views, 175+ upvotes) and maintained the production lifecycle across bug fixes and dataset updates.






