Software Engineer for Humanoid Robot Programming

FRONTIER FORCE TECHNOLOGY PTE LTD; Singapore; On-site; SGD 70,000 - 110,000; Full time

Job summary

FRONTIER FORCE TECHNOLOGY PTE LTD is seeking a Software Engineer for Humanoid Robot Programming to join its factory-lab team in Singapore. The role requires hands-on work with real hardware, integration of ROS 2-based stacks, and development of perception and control pipelines on Ubuntu/Linux.

You will work with Python and C++ on ROS 2, implement Nav2 behavior routes, calibrate sensors, and integrate with factory systems via REST/WebSocket APIs.

Qualifications

• Degree in Robotics, Computer Science, Mechatronics or related field.

Skills

Python

C++

Ubuntu Linux

ROS 2

Nav2

AprilTag

ArUco

Imitation learning

PyTorch

Jetson/TensorRT

Unitree SDK

Unitree SDK2py

Cyclone DDS

Git

Education

Degree in Robotics/CS/Mechatronics

Tools

unitree_sdk2py

Cyclone DDS

Fast DDS

Git

Job description

Software Engineer for Humanoid Robot Programming

Requirements:

• Degree in Robotics, Computer Science, Mechatronics or related field.
• At least 2 years of professional experience programming mobile or legged robots that ran on real hardware - not simulation-only.
• Willing and able to work on-site at the factory lab full-time, including handson time with the robot and factory equipment.
• Strong Python and C++ on Ubuntu Linux; git-based workflow.
• ROS 2 (Humble or newer): nodes, actions, launch files, TF, DDS configuration (Cyclone DDS / Fast DDS).
• Nav2: costmaps, behavior trees, waypoint/fixed-route following, recovery behaviors, indoor localization.
• Fiducial docking: April Tag/ArUco detection, camera calibration, closed-loop terminal alignment.
• Imitation-learning manipulation: training and deploying ACT / Diffusion Policy style policies in PyTorch, teleop data collection, dataset curation, on-robot evaluation.
• REST/WebSocket API integration between factory systems and the robot.
• Practical hardware debugging: joint states, IMU,RGB-D cameras, Wi-Fi reliability, power.
• Strongly preferred Hands-on experience with Unitree G1/H1 and unitree_sdk2 / unitree_sdk2py, including its Python-version and Cyclone DDS constraints.
• Humanoid locomotion tuning and safe load-carrying gait.
• Object classification on-robot (magazine type identification) with YOLO/segmentation or VLM-based methods, deployed on Jetson/TensorRT