Description
The role is responsible for researching and iterating lane-perception algorithms in front-facing camera units to support L2/L2+ assisted-driving functions in mass-production projects. It involves designing lightweight Transformer-based lane-detection models, optimizing them for automotive embedded chips, improving robustness in complex lighting, occlusion, and worn-lane scenarios, and completing model quantization and embedded deployment to balance real-time performance and accuracy. The position requires a master's degree or higher, at least two years of autonomous-driving or ADAS perception-algorithm development experience, knowledge of Transformer/CNN architectures and PyTorch, familiarity with BEV or lane-detection methods, experience with model quantization and pruning, and strong Python and C++ skills.
