Lead Robotics Engineer

Lead Engineer — Teleoperation & Data Capture Systems 

Team: Robotics Division Location: Abu Dhabi, UAE — with regular on-site presence at our Dubai collection facility Reports to: Director of AI & Engineering Type: Full-time, on-site 

The role in one line 

You own the physical end of our robot data business: the robots, the rigs, the sensors, the operators, and the guarantee that every hour we capture is an hour a customer will accept and a model can actually learn from. 

Context 

We operate one of the region's largest teleoperation and egocentric data collection programmes, supplying training data to frontier humanoid robotics companies and national research institutes. We run humanoid fleets under teleoperation, wearable egocentric capture rigs, and a growing network of vendor and in-house collection sites. Our data platform — ingestion, quality gates, provenance verification, LeRobot normalisation, self-service delivery — is already in production and led by a dedicated MLOps lead. 

What we need now is that person's counterpart on the physical side. Today our throughput is capped by robot uptime, rig reliability, and the time it takes to qualify a new capture configuration. We are standing up a second collection facility to roughly double monthly delivered hours. You will be the technical owner of that. 

This is a hands-on senior role with team leadership attached. You will spend real time on the floor with robots, cameras and cabling, not only in design reviews. 

What you will own 

Capture hardware and rig engineering 

• Select, qualify and integrate the full capture stack: egocentric cameras, wearable rigs and smart glasses, depth and stereo modules, VR headsets for teleoperation and head-pose telemetry, hand-tracking and biosignal sensors, on-robot sensor suites.

• Run hardware evaluations against explicit acceptance criteria — field of view, resolution, frame timing stability, exposure behaviour under our lighting conditions, ergonomics over a full operator shift, failure modes. 

• Work with external hardware partners and contract manufacturers on custom capture devices: requirements, design review, pilot builds, qualification, and the decision of when a vendor rig is good enough to scale. 

• Own calibration: camera intrinsics and extrinsics, multi-camera and hand–eye calibration, IMU alignment, and the procedures that let a technician re-calibrate a rig correctly without you in the room.

 

Time synchronisation and data integrity 

• Guarantee temporal alignment across heterogeneous sensors running at different rates — video, high-rate biosignals, joint states, head and hand pose, control commands. - Design the clock discipline: hardware triggering, PTP/NTP strategy, common time base,

 

drift characterisation, and the measurement method that proves alignment rather than assuming it. 

• Diagnose and eliminate sub-frame and tens-of-milliseconds offsets. This is not a theoretical requirement — misalignment here is the single fastest way to make a dataset worthless.

 

Fleet operations and throughput 

• Own uptime and delivered hours for the humanoid teleoperation fleet. Maintenance schedules, spares strategy, wear-part replacement cycles, failure logging, MTBF tracking, and repair turnaround. 

• Set and hit a monthly delivered-hours target. Diagnose the bottleneck when we miss — robot availability, operator capacity, rig failures, or rework from QA rejections. - Build the operator function: recruiting, training curriculum, shift structure, per-operator quality tracking, and the feedback loop that turns a rejection reason into an operator correction.

 

New facility bring-up 

• Take a new collection site from empty warehouse to steady-state production: floor layout and capture cells, network and storage, on-site compute, power, safety and access control, robot commissioning, and the SOPs that let the site run without you. - Do this without interrupting production at the existing site.

 

Capture-to-training handoff 

• Define the on-capture data contract — schemas, topics, naming, metadata, session structure — so that what leaves the warehouse converts cleanly into our training format.

• Co-own quality gates and the reject-reason taxonomy with the MLOps lead, and own everything upstream of ingestion. 

• Translate customer acceptance specifications into capture-floor procedure, and defend our data in technical review when a customer disputes a batch.

 

Team and interfaces 

• Lead a mixed team of engineers, technicians and operators. Task allocation, hiring, technical standards, and priority setting. 

• Act as the technical face of collection operations to customers, research partners, and hardware vendors.

 

What we require 

• 7+ yearsin robotics, hardware systems, or sensor engineering, including at least 2 years on humanoid, manipulation, or wearable capture systems at production scale— not exclusively wheeled autonomous navigation. 

• Demonstrable, hands-on multi-sensor time synchronisationwork. You should be able to walk us through a real synchronisation problem you found, how you measured it, and how you fixed it. 

• Strong practical calibration and sensor integrationbackground: multi-camera systems, depth sensors, IMUs, hardware triggering, USB/GMSL/Ethernet capture pipelines. 

• Working knowledge of humanoid teleoperation: VR-based retargeting, gloves or exoskeletons, whole-body or upper-body controllers, and the practical limits of each. Direct experience with Unitree-class humanoids or equivalent is a strong plus. 

• Fluency with robotics data formats and tooling— ROS 2, rosbag/MCAP, and at least conceptual familiarity with imitation-learning dataset formats such as LeRobot. You should be able to explain what makes an episode trainable, not merely recorded. 

• Python and C++ sufficient to write capture tooling, QA scripts and diagnostics yourself. - You have owned an operational metric— uptime, yield, throughput, delivered hours, cost per unit — and not only an algorithmic one. Tell us the number, the target, and what you did when you missed it. 

• Experience leading a technical team or a shift operation, with accountability for delivery rather than only for your own output. 

• Willing to work on the floor, in a warehouse, in the UAE, on-site.

 

Strong plus 

• Biosignal acquisition (EMG, eye tracking) and its synchronisation challenges.

• Experience supplying data to, or training, VLA / robot foundation models (GR00T, π0, ACT, diffusion policy, or similar). 

• Hardware supplier management across the Gulf and Asia, including import, customs and logistics realities. 

• Fixture, mount and enclosure design; CAD and rapid prototyping. 

• Prior experience standing up a physical production or data-collection facility from scratch.