About us

We want to make intelligent machines something people can count on.

We’re Amit and Eitan, the founders of Physicalfleet. We’ve spent years helping teams operate machines far from the people who built them. Now we’re bringing that experience to Physical AI.

We’ve been here before.

In 2018, we founded Upswift because companies were shipping connected products into the world, but maintaining them after they left the lab was still painfully difficult.

We built a platform to monitor, update, and manage fleets of Linux devices remotely. Our customers used it to operate hundreds of thousands of machines worldwide—from welding machines and coffee roasters to devices watching over swimming pools.

Different products. Different industries. The same expectation: once a machine reaches a customer, it needs to keep working.

That experience taught us what production really means. Connections drop. Hardware wears down. Updates fail halfway through. Devices run different versions, in different environments, often thousands of miles away.

In 2021, JFrog acquired Upswift. We continued building there, bringing our experience in edge devices into a broader software infrastructure platform.

Now we’re starting again.

The machines are changing.

For years, we imagined machines we could explain a job to. Machines that could understand their surroundings, handle something unexpected, and do useful work.

We’re beginning to build them.

Some will be entirely new products. Others will look familiar: a tractor, an industrial arm, an inspection machine, a piece of construction equipment. What changes is the intelligence inside them.

We see enormous possibility in that shift. Machines taking on dangerous inspections. Factory equipment adapting to new tasks. Robots helping with work that is exhausting or difficult to get done.

But getting a machine to do something remarkable once is only the beginning. Getting thousands of them through an ordinary working week takes a different kind of engineering.

The machine outlives the model.

A machine can stay in the field for years while the models, applications, and services running inside it go through many generations.

Someone has to make that relationship work.

Teams need to know what is running on each device, deploy changes carefully, understand unexpected behavior, and recover when something goes wrong.

And a machine can look perfectly healthy as a computer while failing at its job.

A robot puts a part in the wrong bin. Its process is running. Its connection is fine. It reports that the task is complete.

Was it the model? The camera? A motor? A recent update?

Understanding the answer means looking across the whole machine—and having the right information recorded when the failure happens.

That’s why we’re building Physicalfleet.

Physicalfleet is the infrastructure platform for operating Physical AI fleets in production.

We bring fleet management, model observability, remote debugging, and over-the-air updates into one place. We help teams understand their machines, capture what happens in the field, and manage changes across models, applications, and operating systems.

The goal is simple: give the people building intelligent machines the infrastructure to keep improving them after they leave the lab.

We’re building for the ordinary working day.

The models will keep changing. The architecture will evolve. There is still a great deal to figure out.

We want to help build a future where someone can give a machine a job and trust it to get the work done. Where teams can learn from experience across their fleets and turn those lessons into better products.

Most people will never see the infrastructure behind that.

They’ll see a machine arrive, put it to work, and get on with their day.

That’s what we’re here to help make possible.

Amit & Eitan
Co-founders, Physicalfleet