Applied Scientist Intern, Computer Vision

About SafeLab

SafeLab AI Engineering provides custom AI and ML solutions to enterprises worldwide.

What you’ll work on

• Scope incoming problems: define the task (detection, segmentation, keypoints, tracking, reconstruction), the metric (mAP, IoU, HOTA, Chamfer distance), and a realistic target before building.
• Train and fine-tune task-specific models, such as YOLO or DETR-family detectors, Mask R-CNN or U-Net segmentation, and keypoint models, on client datasets that often contain only a few hundred images.
• Combine learned models with classical methods where they work better: line and contour extraction, Hough transforms, homography estimation, camera calibration, multi-view geometry.
• Plan data strategy for low-data settings: annotation protocols, augmentation, synthetic data, active learning.
• Deliver working prototypes with error analysis a client engineer can audit, and export models for deployment (ONNX, TensorRT).

Requirements & Tech Stack

• PhD candidate in Computer Science, Electrical Engineering, or a related field.
• First-author paper at a peer-reviewed CV or ML venue (CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, WACV, BMVC, or similar); please list it on your CV.
• Experience training and evaluating vision models in PyTorch on datasets you collected or labeled.
• Working knowledge of OpenCV and projective geometry.

Preferred

• Experience in document image analysis or vectorization, multi-object tracking, pose estimation, 3D reconstruction, or point clouds.
• Inference optimization experience (quantization, pruning, ONNX Runtime, TensorRT).
• Prior work with industry data or collaborators outside your lab.

Details

• Part-time with no minimum hours; work scales with client projects and fits around your research.
• Remote, or hybrid from our Irvine, CA office.
• Pay: $80 to $120 per hour.