Robot saw
Every camera
Wrist and overhead feeds, synced frame by frame.
e.g. 1080p · 30 fps
Desk open now · Istanbul
The night shift for Europe’s robots.
When a robot gets stuck, a trained Overshift operator takes control in seconds, fixes it by hand and gives it back. Every rescue is recorded, so your robots need us a little less each month.
Control tower · Fleet EU-South
Simulated previewIST 03:12
PICK-01Autonomous
ROTTERDAM
AMR-07Autonomous
LYON
PICK-02Autonomous
ROTTERDAM
PICK-04Human in control
BARCELONA REC
AMR-11Autonomous
BARCELONA
PICK-09Autonomous
MILAN
Simulated feed of the Overshift control tower. Robots, sites and numbers are illustrative.
03:12 The problem
Robots still get stuck. Someone has to be awake when they do.
A clear bag the gripper can’t hold. A pallet left in an aisle. A door that sticks. Small things stop robots every day, and when nobody is watching, the line waits until morning.
2–4%
of driving time Cruise robotaxis needed remote human help, the only figure a robotaxi fleet has ever published (2023).
≈€19k
a month to staff one 24/7 seat in Spain in-house. Five people per seat, before training and turnover. Our estimate from Eurostat wage data.
03:13 How a rescue works
Forty-four seconds, start to finish.
+0:00
Robot asks for help
Your robot detects it is stuck (or your rules decide it is) and raises a flag in the control tower.
+0:06
A human takes control
The next free operator, trained on your robot, claims it. One click, no setup, about 25–60 ms away.
+0:07
Fixed by hand
They drive it with a keyboard, gamepad or leader arm, see every camera, and talk to your floor team if needed.
+0:44
Back to work, and learning
Control returns to autonomy. The rescue is saved as a clean training episode for your next model.
03:14 What every rescue records
Your operator’s hands become your robot’s next lesson.
When an operator fixes a robot with VR hand tracking or a leader arm, every movement is recorded as an action your model can learn from, split into the steps of the fix.
- Hand
- 21 points · 3D
- Wrist
- 6-DoF path
- Steps
- frame-accurate
REC · RESCUE-0412PICK-04 · 03:12:07
ReachGraspLiftPlace
Operator did
Operator commands
Every command sent during the takeover: end-effector pose and gripper width, step by step.
EE pose · gripper width
Labelled
Steps of the fix
Frame-accurate start and end of each step.
reach · grasp · lift · place
Labelled
Outcome
Rescued or not, reviewed before it reaches your training set.
success · failure · QA
03:15 What you get
Software
The control tower.
One screen for your whole fleet. Live cameras, one-click takeover, and a recorder that turns every rescue into training data in the format your models already use. Use it with your own staff, ours, or both.
- · Works with ROS 2
- · Keyboard, gamepad, leader arm
- · Rescue recorder · LeRobot export
- · EU data storage
Tour the software →
People
The operators.
A 24/7 desk in Istanbul, trained on your robots before their first shift. Close enough to Europe for real-time hands-on control.
Meet the desk →
03:16 The data deal
Share the rescues. Pay a third less.
The newest robot models, like Physical Intelligence’s π*0.6, learn fastest from exactly this: a human correcting a robot at the moment it fails. Let us keep an anonymised copy of your rescue recordings and your cover costs about a third less. Prefer to keep everything? That’s fine too. Your data never leaves EU storage either way.
24/7 cover for 20 robots, per month
- One in-house 24/7 seat (estimate)
- ≈ €19,000
- Overshift, you keep all data
- $5,800
- Overshift, rescues shared
- $3,800
03:17 Why Istanbul
Close enough to drive a robot arm by hand.
Watching a robot works from anywhere. Steering one by hand needs a short wire. Istanbul sits a few dozen milliseconds from Europe’s industrial heartland, about a third of the delay from the usual remote-work hubs in India or Mexico.
Istanbul → Frankfurt~25 ms
Istanbul → Barcelona~57 ms
Mumbai → Frankfurt~140 ms
Mexico City → Frankfurt~150 ms
Typical network round trip. Varies by provider and route.
Put a human behind every robot tonight.
We’re taking a small group of European fleets as design partners this winter. Tell us what your robots do and where they get stuck. We’ll show you what cover would look like within a week.