Perception & Vision Engineer

We're building the sensory substrate that lets an advanced humanoid robotics platform understand its environment. As a Robotics Researcher in Perception and Vision, you'll own the pipeline from raw sensor data through object detection, 3D scene understanding and semantic representation, producing the outputs that downstream planning and manipulation systems depend on. Your models run on the robot, in real time, in the real world. Closing the sim-to-real gap isn't someone else's problem here, it's core to this role.

What you'll do

• Design, train and deploy perception systems for object detection, segmentation, depth estimation and 3D scene reconstruction

• Build multi-modal pipelines fusing RGB, depth and inertial data into robust real-time representations

• Develop vision models that transfer reliably from simulation to physical hardware

• Optimise inference pipelines for embedded compute constraints

• Work closely with navigation and manipulation teams on downstream requirements

• Drive systematic hardware evaluation and iterate on failure modes

• Contribute to open-source perception research and tooling

What you'll bring

• Deep foundations in computer vision, 3D geometry and deep learning

• Hands-on experience deploying perception systems on physical robots or real-time embedded platforms

• Strong Python and C++; strong PyTorch or JAX

• A track record of taking perception models from research prototype to deployed inference

• Sensor fusion experience across camera, depth and inertial modalities

Nice to have

• Vision-language models, open-vocabulary detection, or embodied scene understanding

• NeRF, Gaussian splatting, or differentiable rendering

• Manipulation or mobile robotics perception experience

• Publications at CVPR, ICCV, ECCV, CoRL or equivalent