EGOCENTRIC, MULTIMODAL HUMAN DEMONSTRATION DATA · PHYSICAL AI
The data behind Physical AI, and the models built on it.
Humanoid robots have to learn delicate, two-handed work, the kind that can’t be programmed, only learned by doing. Roborecs captures that in the EU, starting with first-person human demonstrations and adding the force and touch a camera can’t see. We capture it to license to robot makers, while building the specialist model it powers next.
first-person sourcestructured training data
THE BOTTLENECK
Robot data can’t be scraped. It has to be captured, one demonstration at a time.
Whoever can capture this data at scale, cheaply, and under clear consent will supply the industry. That is what Roborecs is built to do.
Sources: Scale AI · Bessemer Venture Partners
The data
What egocentric capture is, the full range of tasks it covers, and the research that says it trains robots.
WHAT WE CAPTURE
Egocentric data is what a person sees while doing a task.
Recorded from the first-person viewpoint a robot will actually have. Our capture rig, a 300-degree head-mounted camera set plus a camera on each wrist, records the scene, the hands, the contact, and the order of steps, from the inside of the task rather than from a camera on the wall.
A model learns to recognize the task from the outside.
A model learns to perform it from the inside, the same view the robot has at run time.
The wall camera never sees the grip. The operator’s view does.
Where this data sits in how robots are trained
- Robot teleoperationsmallest, most expensive
- Simulation and syntheticgenerated, mid-scale
- Human egocentric videowidest, cheapest, embodiment-agnosticRoborecs captures here
Foundation models are pretrained on the wide base of human video, then fine-tuned upward. The base is the largest, cheapest, most reusable layer, and the one that cannot be scraped at the fidelity training needs. Framing after NVIDIA Isaac GR00T N1.
The full range of physical work, in first person
First-person demonstrations across the tasks robots must learn, from kitchen benches to workshop tables to industrial floors, recorded through the head-and-wrist rig.
Every session · 7-cam RGB · depth · IMU · audio · pose · spatial trajectories
Bike maintenance 
Produce harvest 
Dumpling folding 
Shrimp peeling 
PCB soldering 
Knife work
Logistics pick-pack 
CNC machining 
Bricklaying 
Bench metalwork 
Quality inspection
Representative capture tasks and sites, illustrative of the diversity the program records.
The case for starting with egocentric data.
01 · WHAT THE RESEARCH SHOWS
External results, not Roborecs measurements.
Robot task success from pretraining on 20,000+ hours of human egocentric video, versus training from scratch.
NVIDIA EgoScaleSuccess across 7 tasks with 20 minutes of human data per task, and zero robot data in training.
EgoZero02 · WHY WE START HERE
For most manipulation tasks, an hour of egocentric capture rivals an hour of robot teleoperation, at a fraction of the cost. The force- and contact-critical work is where the Phase 2 fidelity layer comes in.
Sources: roboticscenter.ai, Unidata, EgoMimic, HumanScale
THE PLATFORM
How the data gets made.
The operation pipeline, the eight synchronized capture channels, the Sofia fidelity facility, and the specialist models the corpus trains.
The company
The moat, the market signal, and the team.
At the crossroads of European deep tech and Asian manufacturing.
Sofia sits between Europe’s deep-tech labs and the Asian robotics supply chain, with the EU’s precision-manufacturing base, electronics and automotive, next door. That is where humanoids deploy first, and the natural market for EU-jurisdiction training data. The Bulgarian cost base is the operational edge; EU jurisdiction is the moat.
PROVENANCE · CONSENT · JURISDICTION
Training data European humanoid makers can lawfully deploy.
Every clip ships with a dataset card and a chain-of-custody. From August 2026 the EU AI Act requires documented provenance and consent for training data. We build it in from the first recording, not bolted on before an audit.
Read the thesis: Physical AI in Europe“We are EU jurisdiction by corporate structure, not just by data location. Under the US CLOUD Act, US-based data suppliers remain subject to US subpoena regardless of where their servers sit. EU-based humanoid customers cannot ignore this.”
EU policy is moving the same way: the June 2026 EU Technological Sovereignty Package and the proposed Cloud and AI Development Act steer public and regulated buyers toward suppliers outside third-country legal reach.
Data residency is not data sovereignty.
A dataset on EU servers is not EU-governed if a non-EU company controls it: the US CLOUD Act can still compel its production. Roborecs holds EU jurisdiction by corporate structure, so consent, provenance, and control stay in Europe end to end.
EU jurisdiction
By corporate structure, not just EU servers. No non-EU parent the CLOUD Act can reach.
Consent and provenance
Documented per session, anonymized at source. GDPR by design, EU AI Act Article 10 aligned.
A sovereign supply
The egocentric demonstration-data layer, captured in Europe and licensable on EU terms.
No open data layer exists yet in Europe. We’re building it.
Embodied AI data infrastructure is the next major investable category, and the market has started funding it. Roborecs is building the European node.
Cumulative humanoid investment surpasses $6B. Foundation model labs start competing for physical-world training data.
Benchmark + Conviction back the household motion-capture model. Gig workers wearing sensor gloves, the US household data layer takes shape.
SoftBank, Bezos Expeditions, NVentures back Pittsburgh’s universal robotics foundation model. Revenue goes $0 → $30M ARR in months.
20+ companies above $1.4B valuation. AgiBot open-sources AGIBOT WORLD 2026. The volume play is firmly Chinese.
Framework Ventures leads the round. Human-motion data capture for robot training is now a standalone fundable category, validated by institutional capital.
INDUSTRY VALIDATION
Europe’s best-funded robotics company already proved the thesis.
NEURA Robotics proved Europe can lead physical AI training. Roborecs scales the supply side, open to every humanoid maker, not vertically integrated into one hardware company.
“NEURA Gyms are physical training facilities where robots practice complex tasks under controlled variability. This generates Physical AI’s scarcest resource: high-quality, real-world training data.”
NEURA Robotics, April 2026
No single OEM can capture the breadth of consented, EU-jurisdiction data a general manipulation policy needs.
Chinese OEMs lead the humanoid volume play and are expanding into Europe, where consented, EU-jurisdiction capture is exactly what a foreign OEM cannot run itself.
Academic research collaborations and pilot programs. Bulgarian-Chinese Chamber pathway to additional Chinese humanoid OEMs.
Humanoid OEMs and VLA labs that train on multimodal physical data. EU industrials via Bosch, Schaeffler, Volkswagen, Siemens as downstream integrators.
Capture the data your robots can’t collect alone.
We build EU-jurisdiction capture programs for humanoid OEMs and VLA labs, vision, proprioception, force-torque, IMU, and audio, in LeRobot-compatible format. Tell us your robot and target tasks, and we’ll scope a pilot.
Start a conversation →The team behind Roborecs.
Serial entrepreneur. Founded Internet Bulgaria 1996. Chairman, Bulgarian-Chinese Chamber of Commerce, a network reaching both Bulgarian partner factories for capture and Chinese humanoid OEMs as buyers.
MSc Robotics & AI, EPFL. Calibration, sensor characterization, and force-controlled manipulation for multimodal data quality.
+ Legal partner, 20-person Sofia firm covering GDPR, IP, EU AI Act compliance from day one.
Join us in building Europe’s data supply for humanoid AI.
We start with first-person human demonstrations, captured in the EU with documented consent and provenance. The fidelity layer comes next, force and touch for contact-rich assembly, and the specialist models the corpus powers after that. It is the EU-jurisdiction, egocentric-first position in a category the market has already funded.
EU AI Act core obligations apply from August 2026. High-risk rules for AI embedded in regulated products, including robots, apply from August 2028 under the 2026 Digital Omnibus. Demand for EU-jurisdiction multimodal training data is accelerating faster than US-based or China-based suppliers can lawfully supply to European humanoid customers.
Request the investor deck.
We share the investor deck and data room with qualified investors. Email us and a founder replies within 24 hours.
Request the deck →FAQ
Questions partners and investors ask.
Who owns the data?
You license a defined, consent-cleared corpus with full provenance. Roborecs retains the right to build its own specialist models on the aggregate. Custom commissions can include full IP assignment.
How is demonstrator consent handled?
Every operator signs a data-use and likeness release before capture. Recordings are provenance-tracked per session and anonymized at source: GDPR by design, aligned with the EU AI Act’s provenance requirements.
What formats do you ship?
LeRobot v3 and HDF5, with sub-millisecond timestamps across channels, compatible with NVIDIA Isaac GR00T N1.7 and Physical Intelligence π0 pipelines. We do not claim any model trains on our data; we ship the format the ecosystem already uses.
Why start with egocentric capture, not teleoperation?
It is the cheapest and most scalable layer, and 2025 to 2026 research (EgoScale, EgoMimic, HumanScale) shows human first-person data can train robots as effectively as teleoperation at a fraction of the cost. The force and touch a camera cannot carry are the Phase-2 fidelity layer.
How are you different from NEURA, a teleoperation-data vendor, or a robot maker collecting its own data?
We start with egocentric human demonstration, where teleop-first vendors start with the most expensive layer. We are a neutral supplier, not a robot maker, so any OEM can license the corpus without feeding a rival’s pipeline. And we capture under EU jurisdiction by corporate structure, with consent and provenance built in from the first recording. NEURA is investing €17M into an in-house facility for its own robots; we build the shared data supply the whole field can license.
When is the facility operational?
Egocentric capture is the phase we start with. The Sofia facility that adds multimodal fidelity is targeted for Q3 2027. We label what ships now versus what comes next, and we do not describe the facility in the present tense.
How does a partner run a pilot?
Tell us your robot and target tasks. We scope a capture program, agree a sample spec, and deliver an evaluation set before any volume commitment. Email [email protected] and a founder replies within 24 hours.