Robotics Perception Engineer – Real-World Vision Systems

Lightspeed; Northbrook (IL); On-site; USD 120,000 - 160,000; Full time

Job summary

LightSpeed Build Technologies seeks a Perception Engineer to own the robot vision stack, calibrate multi-camera systems, and integrate perception with ROS2-based control. You will develop 3D vision, object detection, and pose estimation for on-site manipulation in real-world environments.

Ideal candidates bring 3+ years in robotic perception, Python/C++, and deep learning with PyTorch, plus experience with CUDA and tensorRT for real-time inference.

Qualifications

• 3+ years building and deploying perception systems on real robots or real-time systems.
• Hands-on experience calibrating multi-camera setups (intrinsic/extrinsic/hand-eye) and sensor synchronization.
• Experience with RGB-D and stereo cameras in real deployments.
• Strong fundamentals in camera geometry and 3D reconstruction.
• Deep learning experience with detection/segmentation/pose estimation (PyTorch).

Responsibilities

• Design, calibrate, and deploy multi-camera systems on robot cells and tooling.
• Develop perception to guide manipulation and verify kinematics via vision.
• Localize robots and tools using fiducials and multi-view fusion.
• Build 3D object detection, segmentation, and 6-DoF pose estimation.
• Create point cloud pipelines for manipulation and scene understanding.
• Profile and optimize latency from sensor input to action.

Skills

Perception

Robot vision

Python

C++

PyTorch

Camera calibration

Education

MS or PhD in Computer Vision / Robotics

Tools

ROS2

OpenCV

PCL

CUDA

TensorRT

Job description

LightSpeed Build Technologies seeks a Perception Engineer to own the robot vision stack, calibrate multi-camera systems, and integrate perception with ROS2-based control. You will develop 3D vision, object detection, and pose estimation for on-site manipulation in real-world environments.

Ideal candidates bring 3+ years in robotic perception, Python/C++, and deep learning with PyTorch, plus experience with CUDA and tensorRT for real-time inference.