Case study
Naetha Human: a private atlas of the family's bodies and health
The health history of a whole family pinned to personal 3D models. On its own hardware, with no vendor account.
- Type
- Kol.ski's own product
- For whom
- Families, later also clinics
- Scope
- Concept, server, web and iPhone applications, AI and 3D worker
- Status
- Family version in daily use, concept 1.0
Challenge
The history of a human body is scattered. A photo of a skin change sits in the camera roll with no record of where on the body it was taken. Test results live in laboratory portals, children's measurements in a health booklet, imaging in hospital systems that do not talk to each other.
Since when, and has it changed?
The answer to that question from a doctor is usually a guess. And children's health data is the last thing anyone wants to hand to someone else's servers.
Solution
Every person has their own 3D model, which is a map and an index of their history. A parent holds a finger on the child's model where the change is, takes a photo, saves. Three weeks later the camera shows the previous photo as a ghost to line up the frame, and at the appointment two dated shots sit side by side.
The body as a map
A place on the model leads to the observations, photos, measurements and examinations of that place.
Time as a dimension
Every record has the moment it happened and the moment it was recorded. The time machine overlays two versions of the body.
Data stays with the family
The server, the files and the AI processing run on the family's hardware. The app works offline and syncs in the background.
AI proposes, a human approves
AI organises and points out differences between photos. Every result is a draft for a human to approve.
Result
- 93stable anatomical regions that survive a model replacement
- 4model layers: skin, muscles, skeleton, organs
- 0photos and data on someone else's servers
- 1.0version of the product and platform concept
What works
- The family version in daily use by the founders.
- A body journal operated with one hand, a skin timeline, a time machine and a voice journal.
- The iPhone app works offline and uploads changes in the background.
- Every operation leaves an audit trail, and a whole person can be exported in an open format.
Ahead
- Encryption at rest and the first real scan with a LiDAR phone.
- Import of medical imaging with a slice viewer.
- A clinic as an organisation with time-limited access to a slice of the history.
Naetha Human does not diagnose and does not replace a doctor.
Architecture
Three parts, each of which can run on different hardware. None of them needs a vendor's cloud.
Clients
A native iPhone and iPad application and a web client, with a local cache and an upload queue
- SwiftUI
- RealityKit
- SceneKit
- model-viewer
Organisation server
HTTP API and CLI on one logic layer, database, files, job queue, audit
- Node.js
- Fastify
- PostgreSQL
- OpenAPI
- Cloudflare Access
AI and 3D worker
Processes photos and scans locally, builds model layers from open anatomy atlases
- Python
- Ollama
- Blender
- meshopt
Lessons
- Time needs two stamps. When something happened and when it was recorded: without both, the history lies with every late entry.
- An anchor on the body must not depend on the 3D mesh. Stable regions let us replace the model without losing the history.
- API and CLI on one logic layer. Everything the app does can be run from a script and tested without a browser.
- Children are not scaled-down adults. Fitting the model from measurements has separate rules for each age group.
A similar problem?
We build private, local systems with AI that stay with the owner of the data. Tell us what you want to solve.
The first conversation is free: 30 minutes online, no preparation and no commitment.