Robotics Perception Engineer – Logistics Division
- Hiring from
- Japan
- Work type
- Hybrid
- Posted
626,403 remote jobs, straight from company career pages
100% free · New jobs every minute
Show job descriptionHide job description
The Robotics Perception Engineer position is ideal for a candidate with strong skills in computer vision, deep learning, and 3D perception who enjoys seeing their algorithms run on real robots. The main responsibility will be to develop and improve the perception system that enables our custom-made manipulators to automate logistics operations in real-world warehouses at a high success rate.
You will leverage sensor data (RGB, depth, and others) from robots in operation to develop new perception algorithms and AI models, as well as improve and optimize existing ones. Working within the perception team and alongside the motion planning and system integration teams, you will take features from data collection and prototyping through to stable deployment in production.
Responsibilities
Develop and improve perception modules for logistics automation, including object detection, segmentation, keypoint detection, 3D pose estimation, and 3D reconstruction
Implement perception software in C++ and Python, and integrate it with the robot's control and system software
Collect sensor data from robots in operation, curate datasets, and continuously improve algorithms based on quantitative results
Optimize and compress deployed models for real-time inference on limited compute
Proactively identify and break down issues, investigate logs/metrics, conduct tests, and deliver fixes/features end-to-end
Break down goals into actionable tasks and manage your own timeline to achieve team objectives
Work closely with motion planning, system integration, and hardware engineers so perception and control work together as one product
Analyze field failures and help raise the system's success rate in production, ensuring safety and stability
Requirements
We don't expect any single candidate to meet 100% of the criteria below. If you match most of them and are excited about the role, we encourage you to apply.
Master's degree or above in Computer Science, Robotics, or a relevant field
3+ years of experience developing perception systems or equivalent (including research)
Proficiency in both modern C++ (C++14 or above) and Python
Strong understanding of low-level computer vision, machine learning, and deep learning, with experience implementing state-of-the-art methods from scratch
Hands-on experience in most of the following:
Customized deep learning applications: object detection, segmentation, keypoint detection, 3D pose estimation, or 3D reconstruction
RGB and depth sensors, including multi-sensor setups (e.g., RGB camera + LiDAR)
Model performance optimization, such as network layer optimization and TensorRT / TorchScript
Deep learning and CV libraries such as PyTorch (and wrapped libraries like mmdetection, fastai), OpenCV, Open3D, and Point Cloud Library
ML experiment tooling such as TensorBoard, MLflow, or Weights & Biases
Quick and accurate problem-solving skills, and careful issue monitoring / debugging
Passion to develop a product used in real operations (product mindset)
Good communication skills and proactive collaboration
Business level of English
Desired Skills
Japanese language skill to communicate with internal/external stakeholders
Experience deploying and maintaining perception systems on robots in the real world
Experience with ROS / ROS2
Knowledge of camera calibration and hand-eye calibration
Experience in domain randomization and synthetic data generation using OpenGL, Blender, Isaac Sim, MuJoCo, etc.
Research experience in CV / AI and related robotics fields (publications at CVPR, ICRA, or similar)
Work experience with manipulators, and familiarity with motion planning and execution
Experience with signal processing and network communication (pub/sub, req/rep, UDP, TCP, DDS)
Experience with Docker, CI/CD, and automated deployment of ML models to robots
Work experience in startups