01
Authorized access
We design project-specific access to professional environments and expert workflows, with documented participation, permissions and use restrictions.
Meval AI, Inc.Europe
Real-world data infrastructure for AI
Meval helps AI teams build authorized access to professional environments and expertise in Europe, with rights and provenance designed in from the start.
Current focus: real-world work data for Physical AI and other high-integrity AI data programs.
01The problem
The human data that advanced AI systems need often lives inside professional environments. It cannot simply be scraped, simulated or obtained through generic crowdsourcing. In Europe, rights, privacy, confidentiality and provenance must be addressed before collection becomes a technical problem.
02What we build
02
Rights and provenance
Every collection is designed to carry a record of what was captured, under which permissions, what transformations were applied and what the buyer is licensed to do with it.
03
European data governance
Our rights and provenance framework is being developed with specialist counsel to support GDPR and EU AI Act requirements in cross-border AI projects.
03Early pipeline sample
Early pipeline sample
A controlled egocentric-video workflow and structured action output used to test capture, metadata and provenance processes.
sample_0142.json
{
"clip_id": "mv_sample_0142",
"modality": "egocentric_video",
"environment": "warehouse_pack_station",
"duration_s": 12.4,
"actions": [
{ "t": 0.0, "verb": "reach", "object": "tote" },
{ "t": 1.8, "verb": "grasp", "object": "tote_handle" },
{ "t": 3.2, "verb": "transfer", "object": "tote", "to": "bench" },
{ "t": 7.6, "verb": "release", "object": "tote" }
],
"rights": {
"consent": "documented",
"jurisdiction": "EU",
"license_scope": "model_training"
}
}Early technical sample — not a commercial dataset and not evidence of model performance.
04Where we are
Where we are
Meval is in design-partner discovery. We are developing the European rights and provenance framework with specialist legal counsel and speaking with AI labs, robotics teams and enterprises about the data they cannot obtain through public web data, simulation or generic crowdsourcing.
We are not claiming a catalogue, a scaled network or completed compliance. We are validating what AI teams actually need before building supply.
If you buy, build or evaluate human training data, we would value a 20-minute conversation about what remains difficult.