Wandering Interest
Selected workFleet operations
Project archive

Everyday Robots at X

At X, the support team lived in the space between prototype engineering and the people working with the robots every day. A misplaced cable, a bad calibration, and a software regression could all look like the same problem until we got to the robot.

Robotics workspace
Project imageRobotics workspace
My role
Robotics and mechatronics contractor
When
July 2018 to October 2021
Field work
Deployment support across three Bay Area sites
Project background

What X was trying to build

Everyday Robots was a moonshot for general-purpose helper robots that could operate around people. A factory robot can count on a fixed work cell. A robot in an office has to navigate changing spaces, find objects, and grasp them without everything being lined up for it. The prototypes gave researchers a platform to work on perception, navigation, and manipulation together.

I was a contractor on the team from July 2018 to October 2021. I started in robotics support and grew into mechatronics work as the program moved from early prototypes to newer fleet hardware. The support team sat between the people designing individual subsystems and the operators who saw the whole robot fail or succeed in the room.

X’s Everyday Robots project overview

Behind the project

Fleet operations

What a ticket actually meant

A report usually began with an observation, not a diagnosis: the robot lost a camera, stopped moving correctly, or wouldn’t complete a test. I pulled logs from the onboard Linux systems, checked joints and sensor connections, and reproduced failures on a bench or test stand as needed. I also built or repaired harnesses and ran hands-on LiDAR and stereo-camera calibration. My responsibility was to fix what we could and give the specialist team something concrete when we couldn’t.

Robot hardware detail
Project imageRobot hardware detail

Testing a build before it reached the fleet

A software release could look fine in isolation and still cause trouble on a physical robot. I helped install builds and run scripted motion and sensor checks on the hardware we supported. When a check exposed a regression, I worked with engineers to narrow down what changed and helped restore a known-good release so the team could keep using the fleet while we investigated.

Robot navigating the workspace
Project imageRobot navigating the workspace

Getting a fleet into real spaces

Deployment across three Bay Area sites involved more than moving robots. We had to work through staging, charging, storage, environmental needs, and how a robot would get back to the team when it failed. I helped coordinate those pieces, trained operators, and wrote repair and operating procedures as the hardware and release process changed.

How it fit together

From a fault report to a field-ready robot

  1. Reproduce itOperator notes, onboard logs, and physical checks
  2. Isolate itSeparate sensor, mechanical, and software faults
  3. Send it back outValidate the repair and document what changed