Back to all positions
TAYEE.AI · Careers
Perception Algorithm Engineer
Own the full lifecycle of autonomous-driving perception models: vision obstacle detection, LiDAR obstacle detection, early-fusion perception and depth estimation.
Responsibilities
- Full-lifecycle development of perception models: vision obstacle detection, LiDAR obstacle detection, early-fusion perception and depth estimation
- Lead model selection, training/tuning, systematic evaluation and data-loop construction to continuously improve algorithm performance and engineering iteration
- Support production integration and deployment; own Badcase analysis, issue triage, model fixes, validation and version release
Requirements
- Expert in mainstream BEV perception (BEVFormer etc.); master multi-view image to BEV transformation including intrinsic/extrinsic calibration; familiar with BEV feature extraction, fusion and decoding
- Expert in at least one open-source codebase: OccNet, OccFormer variants, OpenOccupancy, BEVFormer; strong Transformer tuning and distillation skills
- Multi-task detection in BEV space: vehicles, pedestrians, traffic signs; BEV detection head design (Anchor-based and Anchor-free); BEV semantic/instance segmentation and lane detection
- Plus: TensorRT, ONNX Runtime; model quantization (INT8/FP16), pruning, distillation; operator optimization on Horizon J3/J5, NVIDIA Orin (salary negotiable); publications at CVPR/ICCV/ECCV