Machine Learning & Computer Vision Engineer

Role overview Develop vision and machine-learning capabilities for robotic systems using image, video, sensor, and time-series data. The work spans the full experimentation cycle, from identifying data needs and building evaluation pipelines to comparing approaches and integrating selected models into larger software systems.

Responsibilities - Develop image and video understanding workflows for detection, segmentation, keypoints, pose estimation, and tracking. - Analyse actions and multi-step activity over time, including signals from visual and non-visual data. - Build capabilities for failure detection, confidence estimation, and automated verification. - Create and improve data collection, annotation, training, and evaluation pipelines. - Integrate trained models into broader software systems. - Review research, establish baselines, compare model families, investigate failures, and recommend practical approaches.

Requirements - Strong Python and PyTorch experience, with a solid grasp of modern computer-vision and deep-learning architectures. - Experience working with image or video datasets and at least one relevant area such as detection, segmentation, tracking, pose estimation, or action recognition. - Ability to design controlled experiments and choose meaningful evaluation metrics. - Experience with Linux and GPU-based model training. - Ability to translate research papers into working experiments. - Understanding of data collection, labelling, training, validation, and model integration.

Nice to have - Experience with tools or methods such as YOLO, RT-DETR, Mask R-CNN, ViTPose, Keypoint R-CNN, ByteTrack, VideoMAE, Video Swin Transformer, or MS-TCN. - Familiarity with learning-based control, including Behaviour Cloning, Action Chunking with Transformers, or Diffusion Policy. - Background in multimodal learning, synthetic data, domain adaptation, uncertainty estimation, or robotics applications.

The role calls for taking ownership of an experimentation track: defining the problem, identifying data needs, comparing models, and presenting evidence for a recommended solution.