Every project — whether a small pilot or a large-scale program — follows the same ten-step operational workflow, with customer input and defined outputs at every stage.
The customer provides the target data type, intended model use, geography, participant requirements, capture conditions, metadata, annotation and delivery expectations.
We translate the specification into a concrete collection design: participant profiles, environments, devices, activities, sample sizes and metadata schema.
Participants matching the defined demographic, geographic and device criteria are identified and screened for eligibility.
Each participant is presented with a defined-purpose consent process before any data is captured, covering usage, retention and deletion rights.
Collection is carried out according to the approved design — capturing images, video or activity data under the specified conditions.
Raw data is reviewed, filtered and organized — removing invalid, duplicate or non-compliant samples before annotation.
Labels, tags, bounding boxes, keypoints or other annotations are applied according to the customer's schema and guidelines.
The dataset is reviewed against defined quality criteria, including completeness, annotation accuracy and metadata consistency.
A project-specific privacy review confirms consent coverage, data minimization and secure handling before delivery.
The final dataset is packaged and delivered in the agreed format, along with documentation, licensing and metadata files.
Before running the full workflow at scale, we recommend validating the methodology with a small pilot project.