Perception Software Engineer

Meet the team:

We are looking for a highly motivated Perception Software Engineer with a strong background in machine learning to join our ML Perception team and apply your expertise to real-world multi agent autonomous driving and infrastructure challenges to help Glydcars see, and change the way people move around the world. In this role, you will focus on multimodal efficient deep learning for safety critical systems with built in redundancy and robustness. Collaborating with a multidisciplinary team, you will develop state of the art models and algorithms to enhance our perception stack.

Roles & Responsibilities:

• Develop and implement state-of-the-art optimize efficient real time onboard multimodal multitask ML perception models, and finetune to improve performance 
• Apply techniques to optimize model inference running on Jetson, such as quantization, pruning, architecture search, etc. 
• Develop CUDA kernels and TensorRT plugins to perform custom operations and pre-/post-processing
• Research and development in various areas, including multiview sensor fusion for redundancy, robustness, and safety critical operation; and scene understanding with a focus on anomaly detection.
• Design and perform experiments, and share and present findings, including publication.
• Work with multimodal (camera, lidar, radar) datasets from a multiagent autonomous system

Knowledge, Skills and Abilities:

• Bachelor's degree in Computer Science, Computer Engineering, Electrical Engineering, or a related field
• 3 years of professional experience in Machine Learning, preferably working in automotive applications and perception for autonomous driving
• Strong programming skills (Modern C++ and/or Python)
• Familiarity with deep learning frameworks such as Pytorch
• Familiarity with modern deep learning architectures including transformer based architectures
• Excellent communication and interpersonal skills;
• Candidates with experience in optimizing model inference and/or CUDA kernel development are preferred