Recurring, task-specific data collection for robotics teams
A physical-AI roadmap needs a data supply, not a one-off delivery. The enterprise data program runs continuous collection against your task backlog and ships versioned releases on an agreed cadence.
A standing capture capacity against your task backlog
You keep a prioritized list of tasks. We keep the protocols, the contributors and the annotation capacity to work through it.
Prioritized task backlog
Tasks are specified, scoped and queued together. Priorities can be reordered between capture rounds as the model roadmap moves.
Reserved capture capacity
A committed volume of accepted hours per quarter, with contributors and sites recruited ahead of each round rather than after the request lands.
One annotation schema
Every release in the program shares a schema, so datasets from different tasks and different quarters concatenate without rewriting your loader.
Scheduled versioned releases
Releases ship on a fixed cadence with changelogs, so a training run can always be pinned to an exact dataset version.
Evaluation feedback loop
Each round targets what the previous evaluation exposed: a specific subtask, a specific failure mode, a specific environment.
Named technical contact
One person who knows your schema and your protocols, reachable for scoping questions between rounds.
Three ways to engage
Scope decides the model. All three start from the same technical qualification call.
Custom pilot
From approximately €25,000
One scoped task: capture protocol, agreed episode volume, annotation schema, a single versioned release and an evaluation checkpoint.
Indicative only. Final scope and price are set after a technical qualification call.
Enterprise data program
Contact sales
Continuous collection against a task backlog: reserved capacity, shared schema, scheduled releases and a named technical contact.
Priced against committed quarterly volume, schema complexity and licence scope.
Existing dataset licence
Request access
License a released dataset from the catalogue — episodes, manifests, annotations, provenance records and documentation.
Priced per dataset, licence scope and term.
EU-region infrastructure and purpose-bound access
Raw footage is the sensitive part of this business, and it is handled accordingly.
EU-region data infrastructure
Storage and processing run in EU regions. Raw footage does not leave that boundary in the normal course of processing.
Encrypted in transit and at rest
Uploads run over encrypted channels and stored material is encrypted at rest.
Purpose-bound access
Access to raw material is limited to the processing and review roles that need it, granted per project and revoked when the project closes.
Disclosure logging
Every delivery is logged with recipient, scope, dataset version and time.
Defined retention
Raw material carries a defined retention period per project, and contributor or site withdrawal is applied to future releases.
Named sub-processors
The sub-processors involved in storage and processing are named in the data processing agreement.
Designed for GDPR-conscious physical-data collection. Rights, privacy, and provenance are built into the data pipeline.
This describes how the pipeline is designed and operated. It is not a claim of certification, and no dataset is presented as fully anonymous.
Built to survive a data-governance review
The questions a legal or procurement team asks are the questions the pipeline is built to answer.
- Commercial usage licence
- Written terms covering internal training, model deployment, term and territory. Scope is agreed before delivery, not negotiated after it.
- Per-episode provenance
- Capture date range, environment class, protocol version, processing version and annotation schema version travel with every episode.
- Documented rights chain
- Contributor releases and site authorizations are recorded and reviewed per episode before a release is built.
- Immutable versioned releases
- A published release never changes. Corrections and additions ship as a new version with a changelog and a stable identifier.
- Auditable disclosure log
- Who received what, at which dataset version, and when.
- Data processing agreement
- A DPA covering roles, sub-processors, retention and deletion is part of every engagement.
Most technical questions are already answered in the docs
Dataset structure, annotation schema, data formats, versioning, licensing and security are documented in full, with the loader code and manifest excerpts a reviewer will ask for.
Start with a scoped pilot
Bring a task. On the call we scope the capture protocol, the episode volume, the annotation schema and the delivery format.
Pilots typically start from approximately €25,000. This is an indicative figure only—final scope and pricing are set after a technical qualification call.