Machine Learning & Computer Vision Engineer

We are looking for a Machine Learning & Computer Vision Engineer to develop perception and learning systems for robotic applications using image, video, sensor, and time-series data. You will work across the full ML life-cycle: defining data requirements, building datasets and evaluation pipelines, training models, analyzing failure cases, and integrating validated models into production robotics software. The work spans robot perception, human and object understanding, temporal reasoning, and learning-based behavior.

What you will do

• Develop computer-vision and machine-learning systems for robot perception and decision-making.
• Build and evaluate models for detection, segmentation, pose estimation, tracking, video understanding, and sensor-based learning.
• Define data collection, labelling, dataset versioning, and evaluation workflows for real-world robotics problems.
• Diagnose model failures and improve robustness across lighting, viewpoints, environments, hardware, and operational conditions.
• Work with robotics engineers to integrate models into ROS 2-based systems, simulation workflows, and real robot deployments.
• Design clear experiments, compare approaches rigorously, and communicate recommendations backed by evidence.
• Help establish reproducible training, validation, and deployment practices.

What we are looking for

• Strong foundations in machine learning, computer vision, or deep learning.
• Experience building and evaluating models with PyTorch or an equivalent framework.
• Good understanding of data collection, labelling, training, validation, metrics, and model integration.
• Ability to independently own an experimentation track—from problem framing through evaluation and recommendation.
• Strong engineering judgement, clear written communication, and comfort working with ambiguous real-world data.

Good to have

• YOLO, RT-DETR, or Mask R-CNN for detection and segmentation.
• ViTPose, Keypoint R-CNN, or ByteTrack for pose estimation and tracking.
• VideoMAE, Video Swin Transformer, or MS-TCN for video and temporal understanding.
• Behaviour Cloning, Action Chunking with Transformers, or Diffusion Policy for learning-based control.
• Experience with multimodal learning, synthetic data, domain adaptation, uncertainty estimation, or robotics-related ML.