Software Engineer II, Perception

Aurora•Pittsburgh, PA

16h•$126,000 - $201,000•Hybrid

About The Position

Aurora is seeking a software engineer and machine learning practitioner to join its Perception Tracker Team. The role contributes to improving the Aurora Driver’s perception performance and expanding its driverless operational domain. It focuses on end-to-end perception challenges, including tracking dynamic actors, recognizing actions and attributes, and developing semantic understanding of varied traffic environments. The engineer researches and develops machine learning models for challenging real-world conditions, such as highway and urban driving, long-range multi-sensor tracking, varying weather, and degraded sensor conditions. The role also translates insights from on-road and simulated testing into reliable software and takes solutions from prototype through production deployment on the real-time Aurora Driver platform. The position is expected to work from an Aurora office at least three days per week; some roles may require up to five days per week based on operational needs.

Requirements

A master's degree or higher in Computer Science, Robotics, Electrical Engineering, or a related field; alternatively, a bachelor's degree with at least one year of relevant industry software engineering experience.
Proficient software engineering skills in C++ and Python, with an emphasis on clean, scalable, efficient, production-quality code.
Basic knowledge in one or more of deep learning, computer vision, recursive state estimation, structured prediction, or numerical optimization.
Basic knowledge of 3D object-tracking algorithms.
Experience with PyTorch, TensorFlow, or another deep learning framework.

Nice To Haves

Strong understanding of linear algebra, discrete and continuous optimization, supervised and unsupervised methods, generative and discriminative methods, and Hidden Markov Models.
Hands-on experience with vision-language models, foundational models, or world models.
Hands-on experience with CUDA or high-performance GPU programming for accelerated computing.
Experience deploying production-grade computer vision or machine learning models in real-time computing environments.
Demonstrated experience automating machine learning workflows and optimizing training pipelines for efficiency.
Contributions to open-source projects.
Publications at top-tier robotics or computer vision conferences or journals, or strong relevant industry experience in robotics, computer vision, or self-driving technology.
Ability to thrive in highly collaborative, cross-functional engineering environments.

Responsibilities

Address end-to-end perception challenges, including dynamic-actor tracking, action and attribute recognition, and semantic understanding of traffic environments.
Research and develop machine learning models that improve perception in challenging real-world conditions.
Collaborate across teams to refine the perception system and turn on-road and simulated testing insights into robust, well-tested software.
Drive software development from prototype to production, including large-scale data analysis and deployment of production-ready solutions on the real-time Aurora Driver platform.
Contribute to architectural strategy and team collaboration to keep perception systems scalable and efficient.
Design and implement object-tracking algorithms and algorithms that extract, leverage, and fuse information from multiple sensors.
Develop machine learning systems for object tracking and attribute classification.
Build and improve pipelines for large-scale, efficient model training.
Analyze failure modes and implement and test solutions to address identified issues.