About the Role

As a Robotics Engineer (Data & Learning Systems), you will play a key role in building the data and learning infrastructure that power robotics and embodied AI workflows at Kaedim. Working closely with robotics and ML engineers, you’ll enable models to better perceive, understand, and interact with 3D environments by designing high-quality simulation-ready assets, scenes, and the feedback loops that make them better over time.

This role sits between robotics and ML, with a strong focus on the data and asset layer behind perception, simulation, and policy training systems. You'll help bridge the gap between raw 3D and simulation data and the training-ready inputs that downstream embodied AI systems depend on.

Responsibilities

• Design and build the pipelines that turn Kaedim's 3D outputs into simulation-ready assets, including physics properties, articulation, collision meshes, and multi-format export (USD, MJCF, URDF).
• Build evaluation systems that tell us whether a generated asset or scene is actually usable in simulation: does it behave physically, does it train better policies, does it survive domain randomization?
• Extend our synthetic data and scene generation workflows, with a focus on sim-to-real robustness, domain randomization, and embodiment-aware outputs.
• Build tooling for dataset versioning, labeling, and eval tailored to 3D and robotics data.
• Partner with our 3D and ML engineers to integrate data pipelines into both internal generation systems and the formats our customers' training stacks expect.
• Make and own infrastructure decisions (data formats, storage, eval frameworks) that the team will live with.
• Stay up-to-date with advancements in embodied AI literature, and apply them to improve system performance.

Requirements

• Robotics background, with hands-on experience working with real or simulated robotic systems and a working understanding of how training data flows through perception, control, and policy stacks.
• Strong Python, and comfortable with ROS, PyTorch, Tensorflow, or JAX.
• Experience with multi-modal robotics data (images, depth, point clouds, trajectories, sensor logs) and the failure modes that come with it, such as time alignment, silent corruption, and scale.
• Working knowledge of at least one simulator (Isaac Sim/Lab, MuJoCo, Gazebo, Unity/Unreal for robotics) and how assets, scenes, and data flow through it.
• Experience building data pipelines for ML workloads at non-trivial scale, on cloud infrastructure (AWS, GCP, or Azure).
• Strong problem-solving skills and ability to debug complex data and system issues.
• Excellent collaboration and communication skills across software, ML, and robotics teams.
• Pragmatism about scope. You ship usable systems quickly and iterate, rather than building perfect ones slowly.