Responsibilities

1. Camera-hardware selection, debugging, and calibration.

2. Design, development, and optimization of vision algorithms and models for image recognition, object detection, tracking, and segmentation.

3. Development and deployment of deep-learning-based monocular depth estimation and binocular stereo matching.

4. Research and application of 3D human pose estimation and 6D object pose estimation.

5. Tracking cutting-edge deep-learning visual SLAM.

6. Development, debugging, deployment, and delivery of vision algorithms.

Requirements

1. Familiar with camera sensor principles, ISP pipeline, camera parameters and calibration, and image-quality evaluation.

2. Familiar with mainstream network architectures (ResNet, ViT), detection (YOLO, DETR), and segmentation (Mask2Former, SAM).

3. Familiar with at least one monocular depth algorithm (e.g., DepthAnything) and one stereo-matching algorithm (e.g., LightStereo).

4. Knowledge of 3D/6D pose estimation (e.g., PoseFormer, FoundationPose) is a plus.

5. Knowledge of end-to-end visual SLAM (e.g., DROID-SLAM) and Gaussian-Splatting SLAM (e.g., SplaTAM) is a plus.

6. Proficient in PyTorch and common deep-learning models, including implementation, training, tuning, and deployment.

7. Proficient in C++/C and Python with good engineering ability.

8. Full-time bachelor’s or above in computer science, electronic engineering, or related; 2+ years of experience.