Intelligence & Autonomy Careers

A closer look

A robot needs to interpret its surroundings, decide what to do, and turn that decision into controlled motion. This family develops the models and algorithms behind those capabilities.

The work connects learning and perception with geometry, planning, and feedback. A system must be evaluated against the task it performs, including the conditions in which its decisions fail.

What the work involves

Develop representations of the environment, train or implement decision methods, and measure behavior across test conditions. Engineers and researchers examine data quality, uncertainty, task success, timing, and the gap between a controlled test and physical deployment.

IN PRACTICE

Teaching a robot to pick up an object

Perception estimates where the object is. A learned policy or planner selects an approach, and a controller executes the movement. The team tests changes in lighting, object position, and contact to understand where the behavior breaks down.

Explore the professions

REPRESENTATIVE ROLES

Robot Learning Engineer: Develops and evaluates policies that let robots learn actions from demonstrations, experience, or rewards.
Computer Vision Engineer: Builds algorithms that extract objects, features, and spatial information from images and video.
Robotics Perception Engineer: Combines sensor inputs into an estimate of the robot’s surroundings and relevant objects.
Motion Planning Engineer: Develops algorithms that choose feasible paths and trajectories around obstacles and physical limits.
Robotics Controls Engineer: Designs feedback algorithms that turn desired movement into stable, precise machine behavior.
Localization & Mapping Engineer: Builds methods for estimating a robot’s position and constructing maps of its environment.
Autonomy Research Scientist: Investigates new methods for machine decision-making and tests their capabilities and limitations.

Skills & qualifications

Preparation varies between software implementation, controls engineering, and research. Mathematics and programming are useful foundations; research scientist pathways often include graduate study. A single degree requirement would not describe this whole family.

Technical foundations

Python and C++
Linear algebra and probability
Machine learning
Geometry and state estimation
Planning and control
Experimental evaluation

In day-to-day work

Designing reproducible experiments
Interpreting failure cases
Communicating model limitations

How to get started

Implement one bounded task, such as planning a collision-free arm movement in a simulated scene. Record the setup, success criteria, and failure cases. A software engineer can build on programming experience; a controls engineer can start with dynamics and state estimation.

Intelligence & Autonomy