Skip to content

Privacy-first AI for MS management.

MS-Insight finds the links between your symptoms and the weather, using an AI model that runs on your phone. Your health data never leaves it.

Analysing on this device
Network: off

Your last 30 days

Possible triggers, ranked by confidence.

  • Fatigue rises on warmer daysHigh
  • Brain fog follows high humidityMedium
  • Sudden temperature dropsLow

Example output for illustration. MS-Insight supports your decisions. It does not diagnose.

Built for the days you don't have much energy.

On-device AI

Google's Gemma 2B-IT runs directly on your phone through MediaPipe. There is no server to send your logs to, no account to create, and the analysis works the same in airplane mode.

Data sent to the cloud
None
Model size
~2 GB
Works offline
Yes

Symptom and weather correlation

Log fatigue, brain fog, pain and mobility with sliders. Temperature and humidity are captured in the background, so each entry carries its own context.

OCR reports

In development

Photograph an MRI summary or lab result and have the text read on your phone, then added to your timeline next to your symptoms. The scan is processed locally, like everything else.

A language model that doesn't freeze your phone.

Running a 2 GB model on a phone is heavy enough to lock up the interface. MS-Insight moves inference into a separate Dart Isolate, so the screen stays responsive while the model works.

  • Gemma 2B-IT

    Google's open instruction-tuned model, running through MediaPipe. About 125 ms per token in our tests.

  • Dart Isolates

    A worker with its own memory handles inference, away from the UI thread. Peak memory is about 1.9 GB.

  • Privacy-first storage

    Logs live in a local SQLite database inside the app's sandbox. The analysis pipeline makes no network requests.

Built with Flutter. Measured on a mid-range Android 14 phone with 8 GB RAM.

While the model is generating

60 FPS

Smooth. The interface stays interactive.

Measured with
Flutter DevTools

Clearer summaries for your neurologist.

A small on-device model is good at spotting patterns. Writing a careful, well-organised medical summary is a different job. We're exploring the Claude API for that second step.

The idea: after you review it, MS-Insight turns a stripped-down version of your findings into a clear summary you can bring to an appointment. It would be off by default. You would choose what gets shared, see exactly what leaves your phone, and send it only when you tap to send. The on-device analysis would keep working without it.

You choose

Nothing is sent unless you opt in, each time.

You preview

See the exact text that would leave your phone before it does.

You can say no

Every other feature works fully offline without it.

This is a planned direction, not a released feature.

Useful insight doesn't have to cost you your privacy.

Get early access