Atlas Research Engineer, SLAM and Spatial AI

Boston Dynamics; Waltham (MA); On-site; USD 120,000 - 180,000; Full time

Job summary

Boston Dynamics is hiring a research engineer for SLAM and spatial AI on the Atlas VLA research team. The role sits between real-time perception on humanoid hardware and offline reconstruction pipelines used to train and condition learned policies.

This is about giving Atlas a grounded 3D understanding of scenes, not just better image features. You would work on real-time SLAM, 3D reconstruction from logs and multi-view data, and connect geometric scene structure to training workflows for large

Qualifications

• Experience with SLAM, state estimation, 3D reconstruction, or spatial perception.
• Ability to work across robotics, perception, ML, and systems software.
• Comfort testing research code on advanced robotic hardware.
• Interest in connecting classical geometric robotics with large learned models.

Responsibilities

• Build real-time SLAM and perception-based state estimation for Atlas.
• Develop 3D reconstruction pipelines from robot logs, teleoperation data, and multi-view geometry.
• Connect geometric scene structure to VLM and VLA training workflows.
• Prototype spatial AI methods for learned policies.

Skills

SLAM

state estimation

3D reconstruction

spatial perception

robotics

ML

systems software

hardware testing

Tools

GTX?

Job description

What the role actually is

Boston Dynamics is hiring a research engineer for SLAM and spatial AI on the Atlas VLA research team. The role sits between real-time perception on humanoid hardware and offline reconstruction pipelines used to train and condition learned policies.

This is about giving Atlas a grounded 3D understanding of scenes, not just better image features.

What you would work on

• Build real-time SLAM and perception-based state estimation for Atlas
• Develop 3D reconstruction pipelines from robot logs, teleoperation data, and multi-view geometry
• Connect geometric scene structure to VLM and VLA training workflows
• Prototype spatial AI methods that let learned policies reason about space

What they are asking for

• Experience with SLAM, state estimation, 3D reconstruction, or spatial perception
• Ability to work across robotics, perception, ML, and systems software
• Comfort testing research code on advanced robotic hardware
• Interest in connecting classical geometric robotics with large learned models

Why this one is worth a look

Humanoid autonomy will need more than end-to-end pixels if it is going to work in physical spaces people care about. This role is one of the more direct versions of that problem: building the spatial substrate that lets Atlas policies reason about where things actually are.