
Studio XO - Wearable Intelligence
IoT and embedded systems for the fashion tech company behind Lady Gaga's Flying Dress. Fiber optic LEDs, Bluetooth Low Energy, and app-controlled animations — presented at New York Fashion Week.
Physical Intelligence
A workshop in Bangkok with arms, printers, cameras and sensors on the bench. Everything we ship to a factory floor runs here first.
Get in touchFoundation models can see and plan. Getting that onto a machine with no interface, in a plant that cannot stop, is the work nobody demos.
Vision-language-action models are powerful and unpredictable. You cannot ship a robot that works most of the time, so we build the evaluation that says how often, and where it fails.
Between a working model and a working product sit custom firmware, sensor fusion and hardware integration. The most brilliant policy is inert if it cannot talk to a servo.
Two long benches downstairs, an office upstairs looking down through the windows onto the floor. Arms, printers, cameras, sensors, and the cabling that makes them agree with each other. Anything we put on a client’s floor has already failed a few times here.



Live from the bench
The arm takes instructions, moves, and reports what happened. That loop is the whole job: perception in, action out, and a record of both that somebody can audit afterwards.
Custom firmware for edge devices: BLE, WiFi, sensor integration, OTA updates.
Shipped in fiber-optic dresses shown at New York Fashion Week and in industrial storage towers.
Clamp-on current sensing, cameras and edge boxes turn equipment with no telemetry into a live data source. No machine modification, no production stoppage.
Built and wired on our bench, for a cable plant in Thailand where the machines predate any network.
Benchmark pipelines for vision-language-action models, in simulation and on real hardware. Does the robot do what you think it does?
Our own arms are the test rig.
Cameras and detection models watching a process, flagging what a person would flag, at the speed the line runs.
Cameras and a Jetson on a working wire line in Thailand, capturing the footage the detector learns from.
Low-level systems work when the SDK is not enough: direct hardware control, custom protocols, real-time guarantees.
CARB emissions filings captured over an encrypted Bluetooth link to a certified OBD device.
A fault on the floor is only useful once it reaches the system the business runs on. We build that path.
We ran it end to end: a line jam raised as a real SAP maintenance notification and closed there, with SAP computing the downtime. No custom ABAP.
Conferences, partner labs, factory benches in Shenzhen. The robots below are other people’s. Seeing them move in person is how we tell a shipping capability from a good video.




01
We benchmark the hardware and models you already have. What can the system actually do, against what you were told it does?
02
Off-the-shelf parts first: ESP32, Raspberry Pi, Jetson, cameras. A working proof in your hands before anyone commits to custom hardware.
03
Custom firmware, model deployment, sensor fusion, and the path into the system that records the work.
04
Reliability, OTA updates, monitoring, fail-safes. The unglamorous half that separates a demo from a device you trust in the field.

IoT and embedded systems for the fashion tech company behind Lady Gaga's Flying Dress. Fiber optic LEDs, Bluetooth Low Energy, and app-controlled animations — presented at New York Fashion Week.

iPad interface for Apfel's automated storage towers - fully automated high-bay warehouses used in factories across Germany. Real-time inventory management and remote control of the tower's robotic extractor system.

A compliance app for California's Clean Truck Check (CARB). Pairs with a Bluetooth OBD dongle, captures the vehicle emissions snapshot, and files it as a legal record.

A technician app for the trade: search 20,000+ HVAC units, pull the manual and part list, and follow the refrigerant conversion step by step. Native iOS app and API, built end to end.
Bring us the machine nobody can read.
[email protected]