Digital wellness · risk detection

Know your load before
it knows you.

PulseCheck reads the everyday signals — screen time, sleep, stress, movement — and turns them into a real-time risk score, a wellness persona, and the exact next step to take. No account. No judgement. Just a pulse check.

LIVE SIGNALBalanced
How it works

Three inputs. One clear answer.

Everything here runs on a model trained on real check-in data — screen time, sleep, stress, exercise and mood — engineered into a single, explainable risk signal.

01 / SENSE

Read your current state

A short check-in — or a two-minute chat — captures screen time, sleep quality, stress, exercise and mood.

02 / SCORE

Detect the risk

A trained regression model converts your inputs into a 0-10 risk score, and a clustering model places you into one of four wellness personas.

03 / ACT

Get an immediate next step

Breathing, movement, digital-detox or sleep guidance tailored to your biggest risk driver — and a direct line to real support if your score crosses a serious threshold.

Guided check-in

How are you doing right now?

Move the sliders to match how the last few days have actually felt. Your result updates on the right when you run it.

Adjust the sliders and run your check-in to see your risk score, persona and recommendations here.

The model

What's actually running under the hood

A RandomForest regressor predicts the engineered risk score from raw check-in features, and a KMeans model clusters people into four wellness personas. Both are trained on the 500-row digital-wellness dataset shipped in this repo.

0.97
R² (risk model)
0.18
MAE (risk model)
4
wellness personas
500
training rows
PulseCheck guide