Staff Software Engineer, Deep Learning Acceleration

Aurora Innovation•San Francisco, CA

10h•$189,000 - $274,000•Hybrid

About The Position

As a Staff Software Engineer focusing on Deep Learning Acceleration at Aurora, you will play a pivotal role in enhancing the performance of Deep Learning networks utilized in our Autonomous Vehicle (AV) systems. Your primary responsibility will be to conduct thorough performance analysis and optimization of these networks, ensuring they operate efficiently both onboard the vehicle and during training on large-scale data centers. This position requires a deep understanding of software architecture, system performance, and latency issues, as you will be tackling various challenges that arise in these areas. You will collaborate with a team of talented engineers and researchers to develop solutions that improve the overall efficiency and reliability of our self-driving technology. Your work will directly contribute to making transportation safer and more accessible. The role demands a strong analytical mindset, particularly in performance troubleshooting, where you will utilize techniques such as profiling and the roofline model to identify bottlenecks and optimize performance. In addition to your technical skills, you will need to be adaptable and quick to learn new technologies, as the field of deep learning and autonomous systems is rapidly evolving. Strong communication skills are essential, as you will be working in a fast-paced environment with large code bases and collaborating with cross-functional teams.

Requirements

Minimum 5+ years of professional experience in software engineering.
BS, MS, or PhD in Computer Science or a related field.
Strong programming skills in CUDA, C++ and Python
Extensive experience in high-performance computing and parallel programming, specializing in optimizing workloads to reduce GPU memory usage, minimize latency, and/or maximize throughput.
Proficiency in leveraging performance analysis tools such as NVIDIA Nsight Systems , Nsight Compute and applying techniques like roofline model for performance optimization.
Hands-on experience in optimizing DL/ML workloads at the framework level using at least one deep learning framework (e.g., PyTorch, TensorFlow), ensuring efficient and scalable model deployment.
Strong understanding of the fundamentals of computer vision and transformer-based deep learning architectures, with proficiency in foundational neural network building blocks.
Strong analytical skills for diagnosing and troubleshooting performance bottlenecks in complex systems.
Demonstrated ability to quickly learn and adapt to emerging technologies and tools in a fast-paced environment
Experience working on large code bases in a fast-growing environment.
Strong communication skills, enabling effective teamwork across multidisciplinary teams.
Comfortable working in Linux/Unix environments.

Nice To Haves

Hands-on experience in motion planning or related fields such as robotics, autonomous systems, systems software, or computer vision.
Experience with TensorRT, OpenAI Triton, Mojo and other inference acceleration tools.

Responsibilities

Conduct performance analysis and optimization of Deep Learning networks running on the Autonomous Vehicle (AV).
Optimize software architecture, system performance, and latency for deep learning applications.
Work on deployment of deep learning models on the AV and training on large-scale data centers.
Troubleshoot performance issues using profiling and roofline model techniques.
Collaborate with cross-functional teams to enhance the efficiency of self-driving technology.