Platform and AI engine
SCSafe is the name of the platform that runs WellCare AI. It stands for Smart Care Safety AI.
Four layers, from measurement to delivery
This is how a reading from a measuring device is organized, analyzed and delivered to the overview board.
The four layers
| Layer | Role | What it does |
|---|---|---|
| Edge | Measurement and collection | Connects with MFDS-certified measuring devices, wearables and cameras. |
| Stream | Normalization and storage | Stores readings automatically and builds up anonymized data. |
| AI | Analysis | Handled by analysis modules M1–M5. |
| Service | Delivery and action | Delivers results through the overview board, the AI nurse assistant and links to care actions. |
Measurement and interpretation that qualify as medical devices are functions of partner devices certified by Korea's Ministry of Food and Drug Safety (MFDS). SCSafe collects, organizes and analyzes the results.
Analysis modules
The AI layer consists of five modules.
| Module | What it does |
|---|---|
| M1 | Time-series analysis of vital signs, against each person's own baseline |
| M2 | Analysis related to chronic conditions |
| M3·M4 | Responses that combine a generative language model with retrieval |
| M5 | Gait video analysis |
External performance validation (TTA, KAIC) is in progress. We will not describe the modules as validated before the results are out. Performance figures will be published once the metric definitions and source documents are confirmed.
How the analysis works
Each method is explained in plain words, with the technical term next to it.
| In plain words | Technical term | Description |
|---|---|---|
| Compares with the person's usual range | IQR-based outlier analysis | Learns each resident's usual range and looks for readings outside that personal range, not outside an adult average. Example: a systolic reading of 165 for a resident whose usual range is 110–140. |
| Follows the trend | GRU time-series analysis | Looks at how a person's readings have been moving over time. |
| Looks at several measures together | Multivariate analysis | Looks at the relationships between vital signs. |
| Notifies by fixed rules | Notification rules | Staff are notified when readings outside the usual range repeat three or more times and appear in two or more vital signs. |
Notifications follow fixed rules that give the same result every time, not case-by-case judgment.
One analysis engine
One analysis engine covers elderly care, chronic conditions, gait and mobility, and industrial safety.
Each area is at a different stage. WellCare AI for care facilities is in operation. Gait analysis is in development. Use in industrial safety is at the planning stage.
Stages and timing for each area are on the roadmap.
What we disclose and what we do not
We disclose: performance targets, field validation conditions, patent application and registration numbers, and the fact that we follow standards.
We do not disclose: model architecture, preprocessing, data structures, or device-specific internal interface specifications. The interface for connecting with record systems is being developed as a published specification through WellCare Connect.
This protects our intellectual property. We say what the system does, not how it does it. Materials on implementation are available as IR materials after a non-disclosure agreement.
Service notice
WellCare AI is a non-medical health management service that provides reference data for health management and does not replace medical diagnosis or treatment. Measurement and interpretation functions that qualify as medical devices belong to partner devices certified by Korea's Ministry of Food and Drug Safety (MFDS). This system collects, normalizes and analyzes the results of those devices over time, and its use is limited to providing risk information and supporting care work.
Need more technical information?
We provide IR materials for investors and partnership information for technology partners. On the contact form, choose "Investment / IR" or "Partnership".