Who we are and where the data comes from
DataClue is operated by Aurora AI Ventures LLC. We record our own data, check it before it is listed, and sell what the samples show.
Legal entity and business contact
- Legal entity
- Aurora AI Ventures LLC
- Brand
- DataClue (dataclue.app)
- Location
- Jersey City, NJ, United States
- Sales, licensing & billing
- contact@dataclue.app
- Contracting party
- Licence agreements and invoices are issued by Aurora AI Ventures LLC
- Payment
- Stripe invoice, payable by ACH, bank transfer or wire
Procurement teams can request our W-9, a certificate of formation and a mutual NDA by email before evaluation.
Where the data comes from
Every hour in the catalog is recorded first-hand: people performing real tasks in real homes and workplaces, wearing our own capture equipment — GoPro HERO12 and HERO13 cameras with embedded IMU, wearable fisheye cameras, smart glasses and stereo rigs. The data is not scraped from the internet, not synthetic and not resold from third-party collections.
Every recording is made with the written consent of the people filmed. Consent documentation is available to buyers during licence review.
Quality control
- Capture brief. Each scenario has a defined task and environment, and the camera is worn so that both hands and the work surface stay in frame.
- Technical check. Every file is probed for resolution, frame rate, duration and codec. Corrupt, truncated or out-of-spec files are excluded from stock.
- Content review. Sessions are checked for the task actually being performed with hands in view; off-task and idle stretches are not counted as deliverable hours.
- Consent and privacy. Each session is linked to its consent record. Bystanders, screens and identifying marks are screened according to the de-identification terms in your licence.
- Inventory audit. Datasets marked verified have been audited file by file; hours marked indicative are confirmed in writing in your quote before you pay.
Samples match delivered data
- Every public sample is cut from a real session in the same stock we sell, recorded with the same devices and process. Nothing is staged or colour-graded for marketing.
- Specs on dataset pages (resolution, frame rate, duration, device) are measured from the source file, not copied from camera spec sheets.
- Website previews are smaller than the real thing: re-encoded at lower resolution, muted, shortened and cut from sections without bystanders. Delivered files keep the original resolution and frame rate, with IMU where it was captured.
- Every licensed hour ships fully annotated: action segmentation, step-level narration and 2D hand keypoints (3D hand pose on stereo sessions). Annotation is delivered as separate files next to clean video; samples that show keypoints or boxes are visualisations of those files.
- Your licence lists the scenarios, hours and capture specs you are buying. If delivered files don't match them, tell us and we replace them.
The easiest way to check is to try it: download the free sample pack, request evaluation clips with a work email, or order a 10-hour pilot from any dataset page.
Questions? contact@dataclue.app · How licensing works