Bear Robotics

• →None

• Develop vision systems for object detection, segmentation, tracking, pose estimation, depth, scene understanding and task-state recognition.
• Build perception capabilities that support grasping, manipulation, navigation, human awareness and closed-loop robot behaviour.
• Select and adapt modern architectures, including transformer-based, multimodal, open-vocabulary and foundation-model approaches where they provide practical value.

• Create and improve datasets through collection, annotation, synthetic generation, augmentation, active learning and disciplined failure mining.
• Design offline and on-robot evaluation that measures accuracy, calibration, robustness, latency and downstream task impact across representative conditions.

• Integrate cameras and depth sensors; address calibration, synchronisation, coordinate transforms, noise and hardware-specific failure modes.
• Optimise models and pipelines for real-time inference on robot compute while maintaining useful accuracy and predictable resource use.
• Deploy perception components into the robot software stack with strong logging, monitoring, fallback behaviour and regression coverage.

• Diagnose field and lab failures systematically and turn recurring edge cases into data, model or systems improvements.

• Collaborate with robot learning, controls, simulation, hardware and product teams throughout development and pilot validation.

• Performs other duties or takes on specialized responsibilities as assigned.

• Strong practical experience developing computer-vision or multimodal perception systems beyond classroom projects.
• Excellent Python skills and substantial experience with PyTorch or an equivalent deep-learning framework.
• Good knowledge of modern vision methods in several relevant areas, such as detection, segmentation, tracking, pose estimation, depth or representation learning.
• Experience building datasets, training models, selecting metrics and performing structured error analysis.
• Understanding of camera geometry, calibration, coordinate systems and the relationship between perception outputs and physical-world decisions.
• Ability to write production-quality code and integrate models into larger software systems with appropriate tests and observability.
• Strong experimental judgement and the persistence to resolve performance gaps caused by data, models, sensors or deployment conditions.
• Clear communication and effective collaboration across AI, robotics and hardware disciplines.

• Perception for robotic manipulation, mobile robots, autonomous systems or other embodied-AI applications.
• 6D object pose estimation, visual servoing, grasp perception, articulated-object understanding or hand-object interaction.
• RGB-D, stereo, event cameras, LiDAR or multi-camera systems.
• Open-vocabulary perception, vision-language models, self-supervised learning or multimodal foundation models.
• Synthetic data, domain randomisation, simulation and sim-to-real transfer.
• CUDA, TensorRT, ONNX or other inference optimisation and profiling tools.
• ROS/ROS 2 and deployment on embedded or edge GPU hardware.

The physical demands described here are representative of those that must be met by an employee to successfully perform the essential functions of this job. Reasonable accommodations may be made to enable individuals with disabilities to perform the essential functions.

• Prolonged periods of sitting/standing at a desk and working on a computer. The employee routinely is required to sit; stand, walk; talk and hear; use hands to keyboard.
• Specific vision abilities required by this job include close vision, color vision, peripheral vision, depth perception, and ability to adjust focus.
• Ability to lift 30 lbs.
• Occasional travel to partner or pilot sites may be required.

• A degree in computer science, machine learning, robotics, engineering or a related field, or equivalent practical experience.
• At least three years' relevant work experience.