Senior Software Engineer - Perception State Estimation

Relevant knowledge and experience in machine learning, with a proven track record of developing and deploying deep learning solutions using PyTorch or similar frameworks

Experience in developing multi-object tracking systems using classical algorithms or machine learning algorithms

Proven experience in shipping perception software products to industry or consumers

At least 4 years of development experience in Python/C++ environment

with machine learning focus, or equivalent experience

Experience developing and deploying machine learning models with compute constraints

High-quality individual and family medical, dental, and vision insurance

Health savings account with available employer match

Employer-matched 401(k) retirement plan with immediate vesting

Employer-paid group term life insurance and the option to elect voluntary life insurance

Paid parental leave

Paid medical leave

Develop estimation algorithms for road feature estimation such as lane lines and speed limit as well as estimating the road shape.

Develop forecasting algorithms for actors in the scene to estimate time-to-collision and threat levels for automatic emergency braking

Read literature, analyze raw data, and design state-of-the-art solutions

Transition solutions from the lab to the test track and public roads to ensure successful production-level implementation

Collaborate with perception experts and experienced roboticist on algorithm design, prototyping, testing, deployment, and productization

Build and maintain industry-leading software practices and principles

The State Estimation team is a group of highly skilled and experienced professionals who specialize in cutting-edge multi-object tracking, scene estimation, and machine learning technology. Together, we collaborate to create advanced Bayesian filters, graph models, and deep learning models that are capable of temporally tracking both static and dynamic actors as well as estimating road features. The State Estimation team is the interface of the perception system to various downstream autonomy consumers including motion planning, prediction, and localization.

The team's primary focus is on developing compute-efficient models and systems that can perform a wide range of tasks such as closed world multi-object tracking, track-to-detection data association, object motion forecasting, uncertainty estimation, lane line temporal smoothing, perception based road shape generation, and open world scene tracking and state estimation. The ultimate goal is to take these algorithms from the lab to the road, ensuring that they are optimized for onboard performance and able to function as production-grade perception systems on vehicles.

To achieve this goal, the team constantly stays up-to-date with the latest research literature and pushes the boundaries of what is possible. We are dedicated to developing cutting-edge tracking algorithms, ML algorithms, and models that can help vehicles reason about the world around them in real-time.