Robotics Applied Scientist, Spatio-Temporal Scene Understanding & Memory

Description

The Irvine team works on embodied AI with real robots, real sensors and field deployments. The team is based in Southern California's robotics ecosystem and builds risk-aware, reliable, field-ready AI systems for hard robotics problems.

Work here ships, gets tested on hardware and improves through real deployments. The approach goes beyond purely data-driven methods and transformer-only architectures, pairing careful engineering with learning systems that already run in deployed products and improve each time the robots operate in the field.

Responsibilities

• Research a suitable representation for the robot's mission memory: a persistent spatio-temporal record of what was seen, where, when and how it changed across missions and sites, so that moving, vanishing or degrading assets are tracked instead of rediscovered.
• Build world models that let the robot reason and plan over this memory, and recover physical parameters through interaction to support physics-aware digital twins.
• Combine learned models with physically grounded ones, treating geometry and dynamics as ground truth.
• Close the gap between perception and reasoning for long-term autonomy.

Requirements

• A PhD in robotics, computer science, electrical, aerospace or mechanical engineering, or a closely related field.
• Publications at venues such as RSS, CoRL, ICRA, IROS, T-FR, T-RO, CVPR, NeurIPS or similar, on scene understanding, semantic SLAM, lifelong or spatio-temporal mapping, world models or embodied reasoning.
• Hands-on work with vision-language or foundation models for spatial reasoning, memory or long-horizon embodied question answering.
• Solid knowledge of 3D representations such as scene graphs, neural or Gaussian-splat maps and object-centric maps, including the estimation behind them.
• Strong Python, and the ability to run your methods on real robots using C++ and ROS 2.
• A clear view of what a representation is for and how to evaluate it, and the ability to work in a fast-paced setting where research questions come from the needs of the next deployment.

Conditions

Base pay falls within a generous range and depends on the individual's background and experience. Final pay also reflects role scope, job-related knowledge and skills, and the Irvine, California market.

Based in Irvine, California.

Benefits

• Work on real hardware with real sensors, and see systems shipped to operate in unstructured, previously unknown environments worldwide.
• Collaborate with a team whose members have worked at organizations including DeepMind, NASA JPL, Boston Dynamics, NVIDIA, Amazon, Tesla Autopilot, Cruise, Zoox, Toyota Research Institute and SpaceX, with a record of field deployments and strong results in DARPA challenge segments.