New

Synheart's infrastructure is live. On-device inference is free, forever.

Synheart
Woman in a peach shirt, seen in profile

The Human Layer for AI.

Runs on device

Compatible with any device, wearable or not

STATE INTELLIGENCE INFRASTRUCTURE

Synheart turns the signals your tools already produce — code, agents, apps, devices — into contextual state that AI can act on: what the person is working toward, which decisions are binding, what changed, and how sure we are. Start with the work signals you already have.

AS SEEN IN

  • Forbes
  • Yahoo Finance
  • AI News
  • Mila

Three Ways In

One human-state model. Three ways to build on it.

On its own, a read on someone's state is just information. Combined with what your tools already know, it lets them guess less.

  • For developers

    Build it into your app

    Use the SDKs to give your product a read on its users' state through HSI, the open standard. Register your app on the Platform to get started.

  • For teams

    Make your AI tools human-aware

    Synapse gives the AI tools your team already uses your organization's knowledge and, with each person's consent, their working state, so they can guess less.

  • For individuals

    Know yourself, on your terms

    Download Life. It learns your personal patterns on your device, like focus, strain and recovery, and you decide which tools may use them.

What You Can Build

Software that understands the person using it.

Here's what teams are building on Synheart today, from consumer wellness apps to infrastructure integrated directly into other companies' products.

  • Wellness apps

    Track focus, stress, and recovery daily, keeping data on their device.

  • Adaptive AI

    Provide context on stress and mental load for better responses.

  • Focus and attention tools

    Identify focus levels and design the app accordingly.

  • Impact insight tools

    See how a product actually affects the people using it, read directly on their device.

  • Adaptive interfaces

    Create products that detect when users are disengaged and adapt accordingly.

  • Research tools

    Give researchers a real-time, structured way to study attention and stress.

Signal Sources

Begin with a physiological or behavioral signal.

There are two kinds of signals that Synheart's HSI infrastructure can process.

  • Illustration of a human torso with the heart glowing

    Signal from the body

    Physiological signals from any device reading the body watch, ring, patch are standardized by Synheart into a consistent stream.

  • Head wearing a smart headset

    Signal from your devices

    Behavioral signals from digital devices: typing rhythm, scrolling pace, and patterns. All content remains private, focusing on interactions, not actions.

  • WHOOP
  • Garmin
  • WATCH
  • FITBIT
  • Pixel
  • Polar
  • Samsung Galaxy Watch

Compatible with wearable and wellness devices, and the digital devices people already use.

How It Works, Simply

One connection. Everything managed.

Synheart handles the signal processing, the on-device computing with locally inferred models, and the output. You integrate once.

Platform

Where your team manages what you've built: keys, usage, subscriptions, and more, from a detailed developer dashboard that shows your app's data in one place.

Explore Platform

Core SDK

The one thing you actually integrate. Connects to signals, runs locally inferred models on the device, and gives you a clear read.

Explore Core SDK
Round wearable with an orange strap showing the time

Syni LLM

An AI model that knows the person's state and responds accordingly.

Explore Syni LLM

SUPPORTED BY

  • Mila
  • NVIDIA
  • AWS
  • Lambda
  • Anthropic
  • OVHcloud

In Practice

A signal comes in. An answer comes out. Instantly.

There are two kinds of signals that Synheart's HSI infrastructure can process.

PhysiologicalBody

Recovery

78/ 100
Confidence1.00
KinematicMovement

Posture

Sitting
Confidence0.90
DigitalInteraction

Focus quality

76/ 100
Confidence0.85
CognitiveInferred

Focus

72/ 100
Confidence0.82
AffectiveInferred

Arousal

41/ 100
Confidence0.76

These are estimates intended for software to build with. When signal confidence is low, Synheart reports that directly.

Laptop with the Synheart developer console open

For Builders

Up and running in minutes.

Integrate Core SDK, pick a signal source, and read structured output immediately.

  • Works the same across every signal source
  • Runs on the device
  • Delivers a clear answer you can build on right away
Read the docs
Smartwatch showing a score of 87

Privacy, Built In

Your personal data stays on your device.

Personal inference is local by default; any cloud or team features are opt-in and permissioned.

  • Personal signals are processed on the device — the raw data never has to leave.
  • Your app receives an answer, like focus: 74, not the raw signal.
  • No account or identity required for on-device use.
Read the docs

Research & Partners

Backed by science, built in the open.

  • Mila logo

    Mila

    Quebec AI Institute Montreal, Canada

    A Mila startup.

    Synheart is part of Mila's Emerging Startup program, at Quebec AI Institute. Our founder worked in Yoshua Bengio's lab at Mila before starting Synheart.

  • Smartwatch on a wrist showing Calm, 62 bpm, with a heartbeat line leading into it

    On-Device AI

    Real-time inference on the wearable itself

    Built for real time.

    Synheart models are small enough to run on the hardware people already wear. Inference runs on the wearable itself, in real time, with nothing sent to the cloud.

  • University campus buildings in autumn

    Academic Partners

    Collaborative research for real world impact

    Research-grounded.

    We collaborate with universities and labs, always with consented participants and ethics oversight, so our work stays grounded in real-world conditions.

  • Abstract spheres and light, standing for open science

    Open Science

    A more open and human centered future

    Open by default.

    Our papers are public, our SDKs are open source, and our datasets are free on Hugging Face, so anyone can check our work, replicate it, and build on it.

Why We Build This

At Synheart, we build the infrastructure that lets systems understand the people they serve.

And What's Built On It

Real products already running on Synheart.

Human state, read on-device and built into products people use.

THE AGE OF AGENTS

AI is about to act for you. It should know how you're doing first.

Agents now write, schedule, decide, and interrupt on people's behalf. They know the task, but not your organization's history or the person in front of them. Synapse brings both to the AI tools you already run: your organization's knowledge and, with consent, each person's working state.

Start Building

Build something that understands people

Give a product a real, live read on how someone is doing: focus, stress, and recovery, all on their own device.

Start free
Man running

faq

Frequently Asked Questions