Perception Engineer

Key Skills

Apache Hadoop, C Programming, C++, Data science techniques, Java Programming, Machine learning techniques, MapReduce, Python Programming, PyTorch, R Programming, Scala Programming, SQL, TensorFlow

Job Description

Caterpillar’s Integrated Components and Solutions is seeking an experienced Perception Software Engineer who is passionate about delivering autonomous vehicle solutions to production. This person is responsible for developing and demonstrating automation technologies, which are often used in conjunction with information systems, remote control, semi-autonomous, and fully autonomous machine control systems. This is an ideal opportunity for someone with experience in developing perception algorithms and techniques and implementing production quality software in C/C++/Python.

Responsibilities:

• Research and develop perception algorithms for personnel detect, object detection, tracking, classification, terrain mapping, localization and others.
• Architect, Design and Build solutions using Computer Vision and Image Processing techniques for Autonomous Vehicle use cases.
• Benchmark and Tune the Perception/Computer Vison Algorithms and the Hardware Accelerators for performance and quality KPIs.
• Participate in R&D programs through entire lifecycle, from initial inception through production and Research and develop new technology in perception
• Work closely with diverse groups across Caterpillar and contribute towards overall Autonomous Vehicles Architecture taking safety into consideration
• Develop production quality code and own your algorithms from design to production
• Run simulation for perception algorithm development and test various scenarios
• Develop deployable code that will run on resource constrained platforms on mining and construction equipment.

Desired Qualifications:

• BE/ME/PhD in Computer Engineering, Robotics Engineering, Mechatronics, Computer Science, Electrical Engineering, Robotics or a related field.
• Sound knowledge of machine learning/deep learning algorithms, principles and their application
• Experience working with Convolutional Neural Networks.
• Experience with classification, detection, segmentation(semantic/panoptic), localization and tracking.
• Familiarity with the PyTorch deep learning framework.
• Good understanding of object detection and models like YOLO, RCNN, and Mask-RCNN, etc.
• Significant working knowledge of Python, including best practices for testing, documentation, and version control using Git
• Strong programming skills with C/C++ using Linux OS, and middle ware such as ROS/ROS2 and DDS.
• Understanding of graph-based data structures (quadtree, octree, KD-tree, etc.) and Precisely localize the vehicle with localization techniques such as SLAM would be an advantage.
• Experience on programming of Computer vision/perception Systems software, sensor fusion that integrates with perception sensors such as IMU, GPS, Stereo Cameras, Lidar, Radar.
• Proficient in anyone of PyTorch/Tensorflow/keras frameworks in Python/C++
• Good knowledge/experience with CUDA/CuDNN tools.
• Experience on ROS (Rosbag and RViz) for simulation with perception
• Strong math background and strong computational geometry skills
• Knowledge on embedded systems (NVIDIA Jetson, Xavier) & xilinx
• Experience with ADAS would be an advantage.
• Hands-on Experience with MATLab / FPGA / ARM processor would be an advantage
• Knowledge of GPU or parallel computing platforms like CUDA.
• Experience with ARM CPU architecture and hardware considerations (memory controllers, buses, caches, etc.).
• Understanding of principles of RADARs and LiDARS.
• Excellent analytical, communication and interpersonal skills.

Preferred Qualifications:

• Experience with C++ 11/14/17 and python
• Experience in implementing State-Of-The-Art(SOTA) models in practice
• Publications in CVPR/NIPS/ICML/ICLR would be an added advantage.