Job Summary
The Edge Application Developer will build the application layer between on-vehicle machine learning (ML) models and driver-facing features on the Fleet Edge platform, including collision alerts and driver distraction detection. The role focuses on compiling, porting, and optimizing ML models for edge deployment, preprocessing sensor inputs, and developing business logic that combines outputs from multiple concurrent models into actionable, real-time decisions. The position requires close collaboration with platform and operating system teams to ensure reliable data integration and performance within real-time vehicle environments.
Key Responsibilities
• Port and compile ML models developed by the science team for on-vehicle edge deployment.
• Build preprocessing layers to process raw platform inputs, including camera, telematics, and sensor data, for model consumption.
• Develop post-processing and business logic to combine outputs from multiple concurrent models, such as object detection, depth estimation, and trajectory analysis, into a single feature decision.
• Implement confidence filtering and output arbitration logic to suppress low-confidence alerts and prioritize competing driver notifications.
• Integrate model outputs into driver-facing features, including alerts, head-up displays (HUDs), and audible warnings, while meeting real-time performance requirements.
• Collaborate with platform and operating system teams, as well as LiDAR pipeline engineers, to ensure data and timing compatibility.
• Execute development tasks within an established architecture and timeline, adapting quickly to evolving implementation needs.
Required Qualifications
• Strong Python programming skills with hands-on experience deploying and tuning ML models on edge or embedded devices.
• Experience with edge model deployment and optimization frameworks such as ONNX Runtime, TensorRT, or TensorFlow Lite.
• Experience working in resource-constrained computing environments.
• Familiarity with real-time or near-real-time application requirements, including latency management and prioritization or arbitration of concurrent outputs.
• Experience integrating multiple model outputs into unified application logic.
• Ability to work within an existing architecture and implementation plan while managing moderate ambiguity.
Preferred Qualifications
• Experience with automotive, robotics, drone, or other edge-device perception and actuation systems.
• Experience with Advanced Driver Assistance Systems (ADAS) or similar safety-critical, low-latency systems.
• Exposure to sensor fusion or multi-model output arbitration logic.
• Experience with model quantization and compression techniques, such as INT8 calibration, pruning, or distillation, for resource-constrained deployment.
