Description
As Senior World Sensing Engineer, you will own how LightSpeed's robots sense and understand the world. You will architect the sensor array on every robot, from cameras and lidar to force-torque and process sensors, and the pipeline that turns those signals into one trusted world model: where everything really is, what state it is in, and how sure we are.
Sensor Array and Pipeline
- Define the sensor suite for every robot family across factory cells, mobile DEX and roof configurations, and upgrade it when a trade needs a signal we don't yet sense:
- Vision: wrist RGB-D and stereo for close-range manipulation, cell and head cameras for scene coverage, global-shutter cameras for fast motion
- Lidar: 3D lidar for mapping, localization and site safety on mobile robots
- Contact: 6-axis force-torque at each wrist, tactile sensing on grippers and hands
- Proprioception: joint encoders, motor current and IMU on arms, bases and the gantry
- Tool and process: laser line profilers and time-of-flight for surface, gap and depth; tool state, flow and pressure for nailers, pumps and spray
- Site and safety: safety-rated area scanners, and site positioning such as a total station or UWB where lidar alone isn't enough
- Own sensor selection, position and orientation on every robot, for coverage, occlusion and changing light; the application team handles mounting and protection against dust and overspray
- Own time synchronization across every sensor, robot state and teleoperation stream
- Own the calibration strategy (intrinsic, extrinsic, hand-eye, camera-lidar) and keep it reliable through tool changes and daily operation
- Build the sensor data pipeline: capture, streaming, compression and logging at rate, delivered to the world model, teleoperation and robot learning
- Monitor sensor health and detect drift before it affects a task
World Model
- Design the world model: an element graph keyed to BIM, poses with uncertainty, and the state of every element (planned, placed, fastened, verified, failed)
- Register the BIM to the as-built, to tool tolerance, for panels in the factory and for rooms on site
- Build sensor fusion and localization so mobile robots know where they are relative to the building, not just a map
- Keep element state current from perception and from the verification results of every primitive
- Own the world-model interface that the planner, primitives and teaching pipeline use, designed so more than one robot can share it
Perception
- Set the direction for a generalized perception model: one shared model for geometry and categories that learns new features from a few examples
- Build measurement for material state: cable and tape position, mud and surface flatness, foam depth, fastener seating
- Build object detection, segmentation and 6-DoF pose estimation for parts, tools and fasteners in cluttered, confined and poorly lit spaces
- Set accuracy targets tied to task success, and build the benchmarks that measure them
- Deploy perception at the edge within a defined latency budget
Foundation and Skill Pipeline
- Co-own the foundation interfaces with the robot learning and agentic planning principals, and own the world-model interface
- Add the features and sensing each new trade needs as it moves through the skill pipeline
- Measure success by how fast a new trade can be taught
- Decide with the robot learning principal what data is collected and how failures are labelled
Technical Leadership
- Set sensing and perception architecture across both robot families
- Mentor perception and sensing engineers
- Make the system something engineers can teach new trades with, without you in the loop
- Stay hands-on: write production code and test on real robots every week