车载相机感知与部署实习生

Caterpillar Caterpillar · Industrial · Wuxi, Jiangsu

The role focuses on deploying and optimizing camera perception algorithms for autonomous driving on embedded platforms, specifically ARM (NVIDIA Jetson/Orin). Responsibilities include model porting, acceleration using CUDA, TensorRT, and cuDNN, quantization, and tracking cutting-edge research for production feasibility.

What you'd actually do

  1. 负责自动驾驶车载 Camera 传感器相关工作,包括多路摄像头的选型、标定、时间同步与数据采集链路搭建。
  2. 参与基于 Camera 的感知算法(目标检测、车道线检测、BEV 感知、深度估计等)在车端嵌入式平台上的部署与优化。
  3. 在 ARM(NVIDIA Jetson / Orin)平台上完成深度学习模型的移植、加速与功耗优化
  4. 使用 CUDA、TensorRT、cuDNN 等工具进行模型量化(INT8/FP16)、算子开发与推理引擎优化,提升端到端实时性。
  5. 跟踪行业前沿(端到端、BEV、Occupancy、VLM/VLA 等)方向在量产平台上的可行性与部署方案。

Skills

Required

  • C++
  • Python
  • CUDA programming
  • TensorRT
  • INT8 quantization
  • ARM architecture (aarch64)
  • PyTorch
  • TensorFlow
  • ONNX
  • 2D/3D vision perception algorithms (YOLO, CenterPoint, BEVFormer, BEVDet, Occupancy)
  • Linux development environment
  • ROS / CyberRT
  • English literature reading ability

Nice to have

  • Jetson embedded AI platform deployment experience
  • Camera driver debugging
  • V4L2
  • GStreamer
  • OpenCV
  • Model compression and acceleration techniques (pruning, distillation, sparsity, QAT)
  • Automotive-grade SoC (Orin, J5/J6, 8295) heterogeneous computing acceleration
  • Contributions to autonomous driving open-source projects (Apollo, Autoware, OpenPilot)
  • GitHub high-quality open-source projects
  • Kaggle / autonomous driving competition awards

What the JD emphasized

  • ARM(NVIDIA Jetson / Orin)平台上
  • CUDA
  • TensorRT
  • cuDNN
  • 量化(INT8/FP16)
  • 算子开发
  • 推理引擎优化
  • 端到端实时性
  • BEV
  • Occupancy
  • VLM/VLA

Other signals

  • Deploying ML models on embedded platforms
  • Optimizing deep learning models for inference
  • Real-time performance optimization