Software Engineer, Autonomous Vehicles (Motion Planning)

Helm.ai

1d•Hybrid

About The Position

Helm.ai builds AI software for autonomous driving and robotics. Our "Deep Teaching" methodology is uniquely data and capital efficient, allowing us to surpass traditional approaches. Our unsupervised learning software can train neural networks for L2/L3 autonomy, which drastically reduces the need for human annotation and is hardware-agnostic. We work with some of the world's largest automotive manufacturers towards mass production and we've raised over $100M from Honda, Mando, and others to help us scale. We are looking for versatile engineers to shape core components of the next generation of the Helm.ai autonomous stack. No prior experience in autonomous vehicles is required—we prioritize deep expertise in backend systems and a proven track record in high-impact software engineering. In this senior role, you will design, implement, and test safe, efficient, and reliable vehicle software/hardware systems, collaborate with sensor and compute vendors, partner with Helm.ai’s AI teams, and work with customers to advance highway and urban autonomous driving capabilities.

Requirements

MS/PhD in robotics, Computer science, Electrical Engineering, or equivalent industry experience.
Solid C++ for robust, real-time systems, paired with Python for tooling and data analysis.
Deep familiarity with both planning algorithms (search, sampling and optimization-based) and complex decision architectures (like FSMs and behavior trees).
A solid grasp of vehicle kinematics and dynamics to ensure safe, smooth, and realistic motion profiles.
Hands-on experience with ROS/ROS2 or comparable autonomous middleware.

Nice To Haves

Experience shipping a planner on public roads and analyzing real-world disengagement data.
Hands-on experience with tools like CARLA or custom in-house scenario frameworks.

Responsibilities

Architect core behavior planning: Own the logic and decision frameworks that guide the vehicle through complex maneuvers (like unprotected turns, dynamic merges, and yielding).
Optimize trajectories: Develop robust trajectory generation systems that perfectly balance safety, passenger comfort, and vehicle progress.
Ensure verifiable safety: Translate complex traffic rules and operational design domain (ODD) constraints into interpretable, highly safe planning behaviors.
Solve the long-tail: Tackle exciting, real-world edge cases, from unpredictable pedestrians and occlusions to complex construction zones.
Drive cross-functional integration: Collaborate closely to seamlessly connect your planner with perception, prediction, localization, and control modules.
Test and validate: Prove out your solutions through rigorous scenario-based simulations, closed-course trials, and live on-road testing.
Optimize for real-time: Deliver robust, highly deterministic performance tailored for embedded automotive hardware.