Perception Scientist

We seek a Staff Perception Scientist to advance how Physical AI systems understand the world on Arm. You will research and develop perception methods that transform camera, depth, radar, lidar, audio, and other sensor data into reliable representations for robotic and autonomous decision-making.

You will provide scientific and technical leadership from problem formulation and experimentation through optimized on-device deployment. This is a high-impact role at the intersection of computer vision, machine learning, multimodal sensing, robotics, and edge computing.

• Develop innovative methods for detection, segmentation, tracking, localization, scene understanding, and multimodal sensor fusion.
• Formulate research questions, design controlled experiments, and translate promising ideas into robust production algorithms.
• Advance model architectures and representations that remain accurate under motion, occlusion, domain shift, and sensor degradation.

• Define datasets, labeling strategies, metrics, benchmarks, and evaluation protocols that reflect real-world product requirements.
• Build reproducible training and analysis pipelines, investigate failure modes, and communicate findings with quantitative evidence.
• Develop methods for efficient data selection, synthetic data, weak or self-supervision, and continuous model improvement.

• Optimize models and algorithms for latency, power, memory, and accuracy across Arm CPUs and heterogeneous AI accelerators.
• Partner with systems engineers to integrate perception pipelines with sensor I/O, runtimes, robotics middleware, and downstream consumers.
• Create reference implementations and diagnostics that make perception behavior observable, testable, and maintainable on-device.

• Partner with robotics, controls, hardware, product, and customer teams to define perception requirements and system tradeoffs.
• Lead research reviews, mentor scientists and engineers, and guide the technical roadmap for on-device perception.
• Publish, patent, or contribute open work when appropriate, and represent Arm in relevant research and developer communities.

• 6+ years of experience in applied perception, computer vision, machine learning, or a related field.
• Proficiency in Python and C++ with PyTorch, JAX, Tensor Flow, or similar frameworks.
• Strong understanding of:
• Deep learning for visual or multimodal perception, training, evaluation, and failure analysis
• Probability, optimization, 3D geometry, coordinate transforms, calibration, and sensor fusion
• Efficient inference, numerical precision, model compression, and edge deployment constraints
• 3+ years of experience moving perception research into real-time or production systems.
• A record of impact through deployed systems, publications, patents, or widely used research artifacts.
• Ability to design rigorous experiments and explain results to research, engineering, and product audiences.
• Excellent scientific judgment, problem-solving, communication, collaboration, and mentoring skills.

• PhD or equivalent experience in computer vision, machine learning, robotics, computer science, or a related field.
• Experience with 3D perception, SLAM, visual-inertial estimation, tracking, occupancy, or world models.
• Experience with multimodal models, self-supervised learning, domain adaptation, or continual learning.
• Familiarity with synthetic data, simulation, active learning, and dataset curation.
• Experience with quantization, pruning, distillation, compilers, or heterogeneous accelerators.
• Background in robotics, autonomous systems,…