Robotics Perception / Sensors AI Engineer

This is the sensor and perception authority at the top of our robotics solution funnel. When a client requirement comes in, e. g., for egocentric data collection, teleoperation, or field deployment, you are the person who determines what a given sensing setup can actually produce, whether it meets the requirement (sub-centimeter positional accuracy, hand-keypoint fidelity, a usable 3D mesh), and, when it doesn't, what device or configuration will.

You are deliberately device-agnostic: XR headsets (e. g., Pico), optical/mocap suits (e. g., OptiTrack), data gloves, depth cameras, and IMU-based rigs are all fair game. You reason about the feed, the geometry, and the physics behind it rather than any single product. You work shoulder-to-shoulder with Product, Engineering, and Solutions to certify device feeds for pilots before we commit to a client, and you keep scanning the market so we adopt new sensing hardware ahead of demand, not behind it.

Responsibilities:

• Own sensor and perception feasibility for incoming robotics and physical AI requirements given a target outcome, determine which sensors and configurations can meet it, and specify what good looks like at the data level (accuracy, coverage, frame, sync).
• Certify device feeds during pilots to validate whether a device delivers the required signal (sub-centimeter tracking, hand keypoints, 3D mesh) against ground-truth expectations and surface gaps early, before delivery commits.
• Stay device-agnostic across XR headsets, optical/mocap suits, data gloves, depth cameras, and IMU rigs; evaluate and onboard new sensing hardware as it reaches the market and brief the team on what it changes.
• Translate sensor capabilities and limits into recommendations that product, engineering, and solutions can act on and that clients understand.
• Run and interpret sensor calibration and characterization (intrinsic/extrinsic), multi-sensor alignment, drift/noise, and coordinate frames.
• Partner with Robotics Applied AI engineers to turn a validated sensing approach into a working data-collection or perception pipeline.
• Support technical discovery, demos, and proofs-of-concept as the sensing SME alongside solutions leads.
• Maintain clear device profiles, feed specs, and known-limitation notes so the team can reuse your findings rather than re-deriving them.

Requirements:

• 3-5 years hands-on in perception, sensing, or robotics ML work that touched real sensor data, not simulation or literature alone.
• Sensor calibration and characterization performed personally, intrinsic/extrinsic, and multi-sensor alignment with the ability to explain what drifted, how it was detected, and what it cost.
• Fluent 3D geometry transformations, camera models, projection, and coordinate frames reasoned about directly, without reaching for a reference.
• Direct experience with at least two of optical/IR tracking, depth cameras, IMU rigs, optical mocap, and XR headset sensor stacks.
• Strong Python and the judgment to turn a spec sheet or paper into a defensible feasibility answer grounded in the feed and the physics rather than brand preference and expressed as accuracy numbers rather than opinion.
• Communicates a sensing trade-off clearly to an engineer and to a client in the same week, without changing the substance.

Preferred Qualification:

• Perception fundamentals implemented, not only understood: pose estimation, tracking, depth, and keypoint estimation.
• Hands-on with XR/egocentric devices (Pico, Quest, or similar) and their tracking / hand-tracking stacks.
• Experience with optical mocap (OptiTrack, Vicon) or data gloves.
• Exposure to egocentric or teleoperation data collection for robot learning.
• Familiarity with ground-truth rigs and accuracy-verification setups.
• LiDAR / point-cloud processing and sensor fusion.

Nice-to-Have:

• ROS / ROS2 exposure.
• Connections into the research/sensing-hardware ecosystem (e. g., IISc, sensor labs).
• A track record of device profiles, feed specs, or evaluation write-ups that other engineers reused.
• Awareness of the robot-learning and VLA pipelines that consume this sensor data downstream.

Technical Stack:

• Languages: Python (NumPy, OpenCV, Open3D).
• Perception / 3D: camera models, projection, calibration, point-cloud processing.
• Sensing: XR headsets, optical/IR mocap, depth cameras, IMU, data gloves.
• Tooling: calibration toolkits; ROS/ROS2 exposure a plus.
• Core Areas: sensor feasibility, calibration, tracking/keypoints, 3D geometry, and accuracy verification.