State Estimation Engineer

Dyna Robotics • Redwood City, CA

3d

About The Position

As our first dedicated State Estimation Engineer, you will own the estimation systems that enable our robots and data collection tools to perceive their state in dynamic, unstructured environments. You will build sensor fusion pipelines, calibration tools, and diagnostics infrastructure—leveraging classical techniques while exploring learning-based approaches to push the boundaries of what's possible. This is a foundational role with significant ownership and direct impact on developing the most robust robotics foundation model.

Requirements

• 5+ years building state estimation systems on physical robots (not just simulation)
• Deep expertise in sensor fusion, visual or LiDAR-based localization, and filtering/optimization methods
• Strong foundation in probability, linear algebra, optimization, and 3D geometry
• Proficiency in C++ for real-time systems; Python for tooling
• Track record of taking an estimation system from prototype to production deployment

Nice To Haves

• MS or PhD with research focus on state estimation, SLAM, VIO, or sensor fusion
• Experience with manipulation or high-DOF robotic systems
• Hands-on experience with learning-based estimation or differentiable filtering
• Contributions to open-source robotics or publications (ICRA, IROS, RSS)

Responsibilities

• Build core estimation systems
• Design and implement state estimation algorithms (EKF, UKF, factor graphs, optimization-based methods) for localization, pose tracking, and contact estimation
• Develop sensor fusion pipelines integrating IMUs, encoders, force/torque sensors, cameras, and LiDAR
• Own visual state estimation (VIO, visual SLAM) and LiDAR-based localization
• Own the full stack from prototype to production
• Build calibration systems, diagnostics tools, and validation benchmarks
• Own estimation for both production robots and data collection tools
• Push the boundaries
• Decide when to ship a robust classical solution now versus invest in a learned approach
• Experiment with learning-based estimation (learned dynamics, neural filtering) to complement classical methods
• Collaborate with AI and hardware teams on sensor selection and integration

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level