SynHeart
SynHeart
SWIP - Measure what Digital does to you
SWIP

Measure what Digital does to you.

We measure steps. Heart rate. Sleep. Screen time. But we rarely measure the one thing that matters most: How did this digital experience affect my state?

DOWNLOAD SWIPEXPLORE THE PROTOCOL
100%
RUNS FULLY OFFLINE
0 BYTES
RAW DATA LEAVES DEVICE
0-100
SCORE RANGE
REAL-TIME
INFERENCE LATENCY

ON-DEVICE ML

Why On Device Matters.

Most wellness systems rely on cloud processing. Synheart is designed differently: all inference runs offline, from RR interval processing to SWIP scoring.

PRIVACY

Raw physiological data stays on device. Only structured outputs—if the user allows—are ever exported. No central biosignal warehouse, ever.

LATENCY

State inference is real-time. No server round-trips, no delay in session scoring. You get immediate, actionable wellness context.

ROBUSTNESS

Works in airplane mode, low-signal environments, and offline-first architectures. Digital wellness should not depend on cloud uptime.

SESSION LIFECYCLE

Users select which apps to monitor. SWIP handles everything automatically.

  1. 1Tracked app opens → session starts.
  2. 2Offline ML computes MSI windows continuously.
  3. 3SWIP calculates Wellness Impact Score
  4. 4App exits → session closes & profile updates

PER-APP WELLNESS PROFILE

Over time, each monitored app builds a rich physiological context.

Mean Impact
averaged across sessions
Variance
consistency score
Stability
intra-session trend
Time-of-day
circadian pattern

THE SCIENCE

Deterministic. Reproducible.
No black boxes.

SWIP uses a fully specified scoring formula. Given the same inputs, any platform will produce the same output — no hidden heuristics, no silent imputation.

SCORING FORMULA

Users select which apps to monitor. SWIP handles everything automatically.

01 Direction normalization
u_{a,t} = 1 - s_{a,t} (if higher-is-worse axis)
02 Confidence-weighted aggregation
u_a = Σ(c_{a,t} · u_{a,t}) / Σ(c_{a,t})
03 Direction normalization
κ = present_axes / expected_axes
04 Confidence-weighted aggregation
W = 100 · clip(U - κ, 0, 1)
Total324
SESSION TREND
Jan 2024Aug 2024

CROSS-PLATFORM REPRODUCIBILITY

Floating-point arithmetic is not associative. Two platforms can silently diverge. Synheart solves this with:

  • Single portable scoring runtime
  • Fixed iteration order
  • Shipped reference test vectors
  • Explicitly versioned profiles

SCIENTIFIC CREDIBILITY REQUIRES REPRODUCIBILITY.

The same SWIP score is computed identically on iOS, Android, Flutter, and backend systems. Wellness scoring is not a black box — it is a published specification.

IOSANDROIDFLUTTERBACKEND

Monitored app buildsA rich physiological

Start exploring
Person wearing a VR headset, representing physiological monitoring in digital experiences.

A layered system, not a single model.

THE ARCHITECTURE

Synheart is built as five distinct, composable layers each with a clear contract, each independently versioned.

SESSION ENGINE

Detects contextual boundaries such as an app opening and initiates a human state session automatically.

OFFLINE ML INFERENCE

Lightweight on device models compute structured axes.

  • Stress/load
  • Focus/engagement
  • Stability
  • Contextual intensity

HSI — HUMAN STATE INTERFACE

The portable state contract.

  • Time-windowed
  • Normalized to [0, 1]
  • Confidence-aware
  • Versioned
  • Privacy-scoped
Synheart architecture

BUILD WITH SYNHEART.

Open SDKs for human state aware apps.

Integrate Synheart Core SDKs, HSI specification, and the SWIP runtime directly into your app. Available on GitHub.

Dart
Flutter
Android
Kotlin
iOS
Swift

WHAT YOU CAN BUILD

  • Human-state-aware applications
  • Adaptive content systems
  • Wellness-aware games
  • Productivity platforms that respond to regulation state

SYNHEART ON GITHUB

All SDK source code, HSI specifications, SWIP reference implementations, and test vectors are publicly available.

github.com/synheart-ai