InOrbit.AI • XR Design • INTERNSHIP
Managing Robot Fleets in Extended Reality
Timeline
3 months
Role
Product Design Intern
TEAM
XR Developer
Senior XR Developer
Platform
Figma
ShapesXR
Meta Quest
TL;DR
Autonomous robot fleets are managed on the factory floor, but the information about them lives on a dashboard somewhere else in the building. This work was about closing that gap — putting alerts, controls and spatial data in the room with the operator.
In the summer of 2025, I interned with InOrbit.AI, designing and prototyping mixed-reality solutions for their robot operations software, Space Intelligence. Over three months I worked closely with engineering and executive leadership, prototyping in Figma and ShapesXR and testing on the Meta Quest. You can find two such prototypes below.
Context
InOrbit builds robot fleet management software. The team wanted to explore bridging that gap between information and action.
In practice, the people responsible for these robots spend the day walking between the terminal and the floor. An alert comes in, they read it at the desk, then they head out and support the robot as best they can. However, the information stays behind while the person moves.
This is the gap that we were hoping to close with the help of Extended Reality, the work was designed in Figma and Shapes XR while it was tested on the Meta Quest, in collaboration with engineering and execuritve leadership.
Prototype #1
A keep-out area gets drawn by walking its corners, instead of dragging a rectangle across a floorplan.
Zones are the parts of a floor a robot should never enter. On a map you place one and hope the map is current. Here you stand where the boundary belongs and put it there.
Prototype #2
When a robot stops, the building shows you the way to it.
The alert names what failed — pressure, heat, calibration drift — and then lays a route across the floor to the machine. It assumes the person reading it is the person who has to walk over and fix it.
What we learned
Three things worth keeping.
Draw less than the room allows.
A headset will let you put a panel anywhere, and the first instinct is to fill the space. What survived testing was almost nothing — an outline on the floor, a dashed route, one panel that stays where you left it.
Standing in the space is the feature.
Both examples swap an abstraction for the thing itself. You walk the corners of a zone rather than dragging a rectangle over a floorplan you hope is current, and you follow a line to the robot rather than matching an ID to a map.
Comfort is the open question.
Nothing we built answers whether an operator wants a headset on for a whole shift. That is the part we did not reach, and it is the part that decides whether any of it is worth shipping.
Curious to know more?
There is a good deal more behind this than the page shows. Ask me about the constraints, the parts that did not work, or anything else you want to know — I am happy to walk through it.
Grateful to the team at InOrbit.AI for those three months.