ROLE SUMMARY :

We're hiring a Robotics Applied AI Engineer to build and ship perception and robotics ML systems across Physical AI engagements. This is a hands-on builder role : you'll implement CV and perception components, wire up data and calibration pipelines, and take solutions from blueprint to working system - including the messy part where a production data collection runs live at scale and needs constant iteration.

The technical spine is computer vision, perception, and 3D geometry; the surrounding robotics stack - LiDAR, calibration, sim-to-real, VLA models - is where you apply that core to real deployments. You'll work under the technical direction of senior engineers and solutions leads, growing toward owning components end to end. We're looking for someone who writes clean code, ramps fast on new tools, and is genuinely curious about robot learning.

KEY RESPONSIBILITIES :

- Build and iterate on CV and perception components - detection, segmentation, pose estimation - under senior technical direction.

- Implement the 3D geometry underneath - SE(3) transformations, camera models, projection, and coordinate-frame management.

- Run sensor calibration (intrinsic/extrinsic, hand-eye, LiDAR - camera) and basic LiDAR processing - registration, ground segmentation, object detection.

- Build and maintain egocentric data-collection pipelines using cameras, depth sensors, and IMU.

- Take validated blueprints into production and keep them running under live load - support large data-collection ramps (hundreds of users / episodes) with rapid reporting, fixes, and iteration.

- Set up simulation environments and support sim-to-real experiments - domain randomization and gap analysis.

- Integrate and evaluate VLA models; assist with fine-tuning large pretrained models under guidance.

- Build supporting services - REST/gRPC APIs and PostgreSQL - to expose model inference and robotics services.

- Support technical discovery, demos, and proof-of-concepts alongside senior engineers and solutions leads.

- Write clean, tested code and clear documentation; contribute to the team's engineering quality.

MUST HAVE :

Non-negotiable. If a candidate misses one of these, this is not the role for them.

- 2 - 4 years building CV, perception, or robotics ML systems - code that ran somewhere real, not coursework or notebooks alone.

- CV and perception fundamentals implemented and debugged - detection, segmentation, pose estimation.

- Working 3D geometry - SE(3) transformations, camera models, projection, and coordinate-frame management.

- Hands-on exposure to sensor calibration and/or LiDAR processing.

- Strong Python and PyTorch, with a demonstrated instance of turning a research paper into working code.

- Clear communicator who takes senior technical direction well and surfaces blockers early rather than late.

PREFERRED QUALIFICATIONS :

You do not need every skill below - we hire for a strong Must Have core and functional familiarity with the rest.

- Simulation experience - Isaac Sim, MuJoCo, Gazebo, or Unity Robotics Hub.

- Conceptual familiarity with VLA models and modern robot-learning approaches.

- Experience keeping a live system running under load - an at-scale data collection or deployment that needed rapid fixes.

- API development basics (REST/gRPC, Docker) and PostgreSQL.

- ROS / ROS2 exposure.

NICE TO HAVE :

- Hands-on with robot hardware (Franka, UR, Unitree, or similar).

- Fine-tuning or evaluating large pretrained / multimodal models.

- Edge deployment exposure - TensorRT or ONNX.

- Personal projects, open-source work, internships, or coursework in robot learning or CV.

TECHNICAL STACK :

- Languages : Python (PyTorch), C# (Unity)

- Perception / 3D : OpenCV, Open3D, PCL, point-cloud processing

- Simulation : Isaac Sim, MuJoCo, Gazebo, Unity Robotics Hub

- Robotics Middleware : ROS / ROS2

- Models : VLA models (RT-2, OpenVLA, Octo), ViTs, multimodal, hosted and open-source

- Deployment : REST/gRPC, Docker, PostgreSQL; TensorRT / ONNX for edge

- Core Areas : CV and perception, 3D geometry, calibration, LiDAR, sensor fusion, sim-to-real

SUCCESS MEASURES :

- Perception and robotics components shipped to spec, well-tested, and cleanly integrated.

- Reliable calibration, data collection, and sim setup that others can build on.

- Fast ramp on new sensors, models, and tools as engagements require.

- Growing independence - moving from guided tasks toward owning components end to end.

- Clean, documented code that raises the team's baseline quality