Parking Perception DNN Engineer

Intelligent machines powered by Artificial Intelligence computers that can learn, reason and interact with people are no longer science fiction. GPU Deep Learning has provided the foundation for machines to learn, perceive, reason and solve problems. Now, NVIDIA’s GPU runs Deep Learning algorithms, simulating human intelligence, and acts as the brain of computers, robots and self-driving cars that can perceive and understand the world.

We are now looking for an extraordinary Senior Perception Engineer to develop and productize NVIDIA’s autonomous driving solutions. As a member of our perception team, you will work on building world-class 3D obstacle perception solutions based on multi-sensor fusion, including cameras, ultrasonic sensors, and radar, to estimate high-resolution reconstruction of the world, such as occupancy networks. The primary approach will be deep learning. You will be challenged to improve robustness and accuracy as well as efficiency of the solutions to fully enable autonomous driving anywhere and anytime.

What you’ll be doing:

• Perception experts with application focus will be on multi-sensor fusion based deep learning model development for obstacle perception/fusion in complex driving environments.

• Applied research and development of innovative deep learning and multi-sensor fusion algorithms to improve output accuracy of 3D obstacle perception solutions under challenging and diverse scenarios, with a focus on high-resolution world reconstruction (e.g., occupancy networks).

• Identify and analyze the strength and weakness of the developed 3D obstacle perception solutions using large scale benchmark data (both real and synthetic) and improve them iteratively through KPI building and optimization. This includes careful data verification, model architecture design, understanding details of loss function engineering, and being capable of finding detailed ML bugs and iterating toward perfection.

• Productize the developed 3D obstacle perception solutions by meeting product requirements for safety, latency, and SW robustness, with a strong emphasis on production deep learning model development.

• Drive and prioritize data-driven development by working with large data collection and labeling teams to bring in high value data to improve perception system accuracy. Efforts will include data collection prioritization and planning, labeling prioritization, so that value of data is maximized.

What we need to see:

• 10+ years of hands-on work experience in developing deep learning and algorithms to solve sophisticated real world problems, and proficiency in using deep learning frameworks (e.g., PyTorch).

• Experience in multi-sensor fusion (cameras, ultrasonic sensors, radar) for perception tasks, particularly in high-resolution world reconstruction.

• Proven experience in production deep learning model development, including careful data verification, model architecture design, loss function engineering, and debugging ML models.

• Experience in data-driven development and collaboration with data and ground truth teams.

• Strong programming skills in python and/or C++.

• Outstanding communication and teamwork skills as we work as a tightly-knit team, always discussing and learning from each other.

• BS/MS/PhD in CS, EE, sciences or related fields (or equivalent experience)

Ways to stand out from the crowd:

• Experience on end-to-end deep learning model development is a plus.

• Proven expertise in developing perception solutions for autonomous driving or robotics using deep learning with multi-sensor input.

• Hands-on experience in developing and deploying DNN-based solutions to embedded platforms for real time applications.

• Good understanding of fundamentals of 3D computer vision, camera calibrations including intrinsic and extrinsic, and sensor fusion principles.

• Experience with development in CUDA language. The ability to implement CUDA kernels as part of training or inference pipelines.