Perception Engineer/Scientist – Foundation Model

• Architect and optimize single-model or multi-task E2E perception networks, from data ingestion through decision output
• Unify object detection, segmentation, depth estimation, and other tasks under a BEV framework
• Implement multi-view projection, feature fusion, and visualization in a BEV representation
• Research and deliver efficient BEV-based tracking, mapping, and behavior prediction modules
• Fuse camera, radar, IMU, and ultrasonic inputs into a cohesive BEV map
• Design multi-modal attention and temporal association strategies to enhance robustness
• Apply quantization, pruning, and distillation to tailor networks for in-house ASIC
• Continuously refine inference latency, compute load, and power consumption to meet automotive real-time requirements
• Survey the latest BEV/E2E papers, propose improvements, and drive prototypes
• Contribute to technical reports, patent filings, and peer-reviewed conference or journal publications

• MS or PhD in Computer Science, Electrical Engineering, or a related field
• 2+ years of experience in computer vision or deep learning R&D
• Proficient in PyTorch and capable of building models from scratch
• Familiar with multi-view geometry and BEV projection
• Experienced with end-to-end training pipelines
• Hands-on experience with model compression on ASIC (quantization, pruning, distillation)
• Strong foundation in detection, segmentation, depth estimation, or tracking
• Knowledge of spatio-temporal synchronization or multi-sensor fusion
• Proficient in Python
• C++ experience is a plus
• Comfortable with Git, Docker, and collaborative development tools
• Strong communication and cross-team collaboration skills
• Passionate about AI and autonomous driving technologies

C++ experience is a plus