Smart gloves, vision glasses and future sensors produce synchronized motion, touch and environment signals.
Building the Data Factory for Physical AI
The Data FactoryFor Physical AI
Arcomni Harbor unifies capture hardware, an iPad edge studio and a customer-owned embodied data harbor for physical AI teams.
One system, end to end
A data production line, from the hand to the cloud.
awear devices capture real-world motion, touch and first-person vision. awear Studio handles calibration, labeling, preprocessing and teleoperation. Arcomni Harbor then caches customer-owned data as indexed warehouse assets for Physical AI training pipelines.
How to read this architecture
The platform separates data production from customer-side robot training.
awear captures signals in the real world, Studio turns those signals into reviewed and connected episodes, and Arcomni Harbor stores the resulting assets as a customer-owned data warehouse. Customers then use governed data for preparation, model training, policy learning and sim-to-real workflows.
Desktop and iPad Studio support calibration, annotation, preprocessing, review and robot teleoperation.
Arcomni Harbor organizes data by project, action, scenario and model need while keeping customer ownership explicit.
Customer teams use the governed warehouse for data preparation, model training, policy learning and sim-to-real deployment.
Capture hardware
One capture format across gloves and vision devices.
Start with the smart glove, add tactile arrays when force matters, and sync vision devices without changing the dataset pipeline.
Both glove SKUs share one calibration flow and one upload format, so mixed fleets do not require separate pipelines.
Edge app
Process the data where it is captured.
awear Studio Desktop handles richer calibration, annotation and robot teleoperation. awear Studio for iPad keeps capture, replay, approval and upload close to operators in the field.
Cloud OS · Custody
Your embodied data harbor. Kept under your control.
Arcomni Harbor versions, indexes and guards the dataset warehouse. It is never ours to keep, sell or train on by default.
Security and ownership
Built for enterprise data custody.
No default training on customer data
AI labels and training exports run inside explicit workspace authorization.
Workspace isolation
Object keys, API keys, metadata and audit events are scoped to each organization.
Provider separation
Stripe, IAP and enterprise contracts produce billing events; product access is decided by entitlement snapshots.
In the field
Teams shipping robots run on Arcomni Harbor.
Company
Built from Singapore for regional data custody.
Arcomni Pte. Ltd. operates Arcomni Harbor as a customer-owned data platform for Physical AI teams, with residency and audit boundaries designed into the product.
Get started
Capture your first episode this week.
Get gloves on operators, process on the iPad, and keep every byte in a cloud the customer owns.