AI Engineer (SLAM)

Foundation•San Francisco, CA

31d

About The Position

A humanoid operating in unstructured, real-world environments must know exactly where it is and what surrounds it at all times — SLAM is the foundational system that makes that possible Our humanoid needs to build and maintain accurate, real-time maps of dynamic environments while simultaneously localizing itself with centimeter-level precision — a problem that requires a dedicated, senior-level focus Off-the-shelf SLAM solutions are not sufficient for a full-body humanoid operating across floors, stairs, cluttered rooms, and outdoor terrain — we need custom, embodied solutions built from the ground up As we move from controlled lab settings to real-world deployment, robust localization and mapping becomes a hard dependency for every upstream system — perception, planning, and manipulation all rely on it We are scaling our autonomy stack and SLAM is the critical infrastructure layer that must be production-ready before the rest of the system can follow This role directly impacts how our humanoid understands its place in the physical world — it is foundational, not peripheral

Requirements

Deep knowledge in vision for robotic systems
Hands-on experience implementing SLAM pipelines in C++ and Python — you have built and tuned these systems end-to-end, not just integrated existing libraries
Strong working knowledge of modern SLAM frameworks: ORB-SLAM3, RTAB-Map, Cartographer, LIO-SAM, or KISS-ICP — and the ability to extend or rewrite core components when needed
Experience with neural or learned SLAM approaches (DROID-SLAM, iMAP, NeRF-SLAM)
Experience with legged or humanoid-specific odometry challenges
Comfortable with probabilistic state estimation, Kalman filtering (EKF/UKF), and particle filters as they apply to real-time localization under uncertainty
Familiar with loop closure detection methods, place recognition networks and strategies for long-term map consistency in changing environments
Hands-on experience with simulation environments such as Isaac Lab, MuJoCo for development, testing, and sim-to-real validation