Robotics Engineer – Imitation Learning & Autonomous Manipulation

About the role

FacilityOps.AI is looking for a Robotics Engineer to develop imitation‑learning and autonomous manipulation solutions for fiber‑optic connector inspection and cleaning in AI data centers. The role is based in Kenya (preferably Nairobi) and combines research, software development and hands‑on testing on real robot arms.

Key responsibilities

• Design and train manipulation policies from human demonstrations using techniques such as ACT, Diffusion Policy or vision‑language‑action models.
• Define data‑collection pipelines for tele‑operated demos, clean, label and version the datasets.
• Integrate learned policies with classical motion‑planning (MoveIt 2), grasp and pose estimation, and force/torque‑aware insertion.
• Build and maintain simulated environments (MuJoCo or Isaac Sim) for testing, data augmentation and sim‑to‑real transfer.
• Implement safety and reliability measures, monitor success rates over hundreds of trials and report results.
• Package policies as ROS 2 nodes, deploy them on robot hardware and debug remotely with the US lead.
• Conduct local testing on a low‑cost arm shipped to Kenya before moving to the Kinova Gen3 platform.

Required profile

• Degree in robotics, mechatronics, mechanical, electrical, computer engineering or computer science, or equivalent hands‑on experience.
• At least 2 years of experience building robot manipulation systems, including a complete data‑to‑policy pipeline on a real arm.
• Hands‑on experience with imitation‑learning and behavior‑cloning policies implemented in PyTorch.

Required skills

• ROS 2 (nodes, topics, actions, TF, launch files) and MoveIt 2.
• Strong Python programming; working knowledge of C++.
• Robot kinematics, coordinate‑frame handling and camera calibration.
• RGB‑D perception for grasping, point‑cloud processing and OpenCV.
• Linux, Git, Docker and training on cloud GPUs.
• Experience with simulation tools such as MuJoCo or Isaac Sim and sim‑to‑real transfer.
• Familiarity with Kinova Gen3, Hello Robot Stretch and LeRobot SO‑101 hardware.