Perception & State Estimation Engineer - DoorDash Air

DoorDash•San Francisco, CA

About The Position

DoorDash Labs is an independent team within DoorDash that explores robotics and automation for the future of last-mile logistics. The Perception & State Estimation Engineer will join the Autonomy Software team and own the development of perception and state estimation for the package pickup and delivery system. The role will influence early system-design decisions, including sensor selection, sensing architecture, and observability analysis. The engineer will develop and tune the drone’s state-estimation stack, including PX4 EKF modifications, fault tolerance, and fusion of multiple sensors with visual-inertial and laser odometry. They will also own estimation and perception for the delivery mechanism, working across dynamics modeling, observation-model design, and sensor selection. The role collaborates with controls, planning, and hardware teams, and includes designing and executing flight tests to validate estimator performance and diagnosing failure modes.

Requirements

Deep experience with state estimation and filtering, including EKF, UKF, MSCKF, or particle filters.
Experience with optimization-based methods such as factor graphs, GTSAM, Ceres, or SymForce.
Experience with visual-inertial odometry (VIO) and/or laser odometry (LIO).
Proficiency in modern computer vision, including feature extraction, tracking, and depth estimation.
Strong C++ and/or Rust skills, with Python for prototyping and analysis.
Experience deploying estimation on real hardware with real sensor noise.
Ability to derive an EKF from scratch, including process and measurement models, observability analysis, and consistency checks.
Deep linear algebra knowledge, including matrix decompositions, sparsity exploitation, and numerical conditioning.
Experience closing a perception-to-control loop on real hardware under degraded conditions.
Comfort working at the intersection of theory and embedded reality.
Passion for delivering outstanding customer experiences and a proactive approach to development, including crafting new tools.
MS or PhD in Computer Science, Electrical Engineering, Mechanical Engineering, Robotics, or a related technical field.

Nice To Haves

Experience with PX4 or ArduPilot EKF internals.
Experience with multi-body state estimation, such as tethered or suspended payloads.
Familiarity with SLAM in GPS-denied environments.
Publications or open-source contributions in state estimation or VIO.
Experience debugging estimator failures in the field rather than only in simulation.

Responsibilities

Own tuning and implementation of the drone state-estimation stack, including PX4 EKF modifications, fault tolerance, and multi-sensor fusion with VIO and laser odometry.
Own state estimation and perception for the delivery mechanism, including dynamics modeling, observation-model design, and sensor selection.
Work with controls and planning teams to close the loop on automated pickup and delivery.
Work with the hardware team to define sensing principles and sensor placement.
Design and execute flight-test campaigns to validate estimator performance.
Identify and resolve estimation failure modes, including degenerate geometries, sensor dropout, and initialization issues.
Influence system-design decisions, including sensor selection, sensing architecture, and observability analysis.