Everything you need to know about Pharo
Getting Started
What does Pharo do?
Pharo trains a personal AI model unique to your body. You label recurring experiences you want to track—flares, migraines, bad days, whatever matters to you—and Pharo identifies which vital sign changes precede those patterns. Your model then runs directly on your phone, giving you a daily risk score so you can see patterns coming before you feel them.
What data does Pharo need?
Pharo reads resting heart rate, heart rate variability, respiratory rate, sleep, and activity data from Apple Health. You need an Apple Watch or compatible device that records these metrics. The more consistent your data history, the better your model will be.
How do I get started?
Download the app, connect Apple Health, and label past episodes. Pharo is built for retrospective labeling—you don't need to wait for new episodes to happen. Think back to when your pattern occurred (flares, migraines, bad stretches), mark those dates, and once you've labeled at least 5 episodes you can train your model right away. If you have months of Apple Health data already on your phone, you could go from download to a working personal model in minutes.
How It Works
What's a "label"?
A label is a set of days marking when a pattern occurred—a migraine episode, a flare period, an anxiety spike, a stretch of bad sleep. You name the pattern anything you want and mark the days it happened. Pharo uses these labels to learn what your vitals look like before, during, and after your episodes.
What happens when I train my model?
When you hit train, Pharo sends your health data window and labels to our servers where an AI-driven agent takes over. This agent runs a sophisticated series of machine learning experiments—testing different models, tuning parameters, and evaluating results—until it finds the best model for your specific data. It then produces a compact CoreML model that's downloaded directly to your device. After that, your data is deleted from our servers and all predictions run locally on your phone.
What makes the training so special?
Most health apps use one-size-fits-all algorithms. Pharo's training is powered by an AI agent that acts like a data scientist dedicated to your data. It autonomously designs and runs dozens of ML experiments, analyzes results, adjusts its approach, and iterates until it finds the model that works best for your unique physiology. It's personalized machine learning that would normally require a team of engineers.
How do predictions work after training?
Your trained model lives on your device and runs entirely offline. Each day, it analyzes your recent vital signs and produces a risk score. You can check your daily score and view your full history of risk scores over time to spot trends. No internet connection required for predictions.
How early can it predict patterns?
It depends on your body and pattern, but Pharo often detects changes in vital signs 2-3 weeks before symptoms appear. Your body starts shifting before you notice anything—subtle changes in heart rate, respiratory rate, sleep quality—and Pharo's model is trained to spot exactly those shifts in your data. The more historical data and labels you have, the better the model can learn your patterns.
Privacy & Data
Is my health data safe?
Yes. During training, your health data and labels are sent to our servers, but they are stored ephemerally—used only for the duration of model training and then deleted. Once your CoreML model is built and downloaded to your device, your data is removed from our servers. After training, all inference runs entirely on your device with no server communication.
Do you sell my data?
No. We will never sell, share, or monetize your health data. Our business model is subscriptions, not data harvesting. Your data exists on our servers only for the minutes it takes to train your model, then it's gone.
Does my model need internet to work?
No. Once trained and downloaded, your CoreML model runs entirely on your iPhone. Daily risk scores are computed locally with no network calls. You only need an internet connection when training a new model.
Troubleshooting
Why can't I train yet?
You need at least 5 labeled episodes before training. Each label should mark the days when your pattern occurred. If you have enough labels but training still isn't available, make sure you have sufficient health data history from Apple Health.
Why isn't my data showing?
Check that you've granted HealthKit permissions in Settings → Privacy & Security → Health → Pharo. Also ensure your Apple Watch is syncing data to your iPhone regularly.
Training failed—what now?
This usually means there isn't enough data variety for the model to learn from. Make sure you have a good stretch of health data (30+ days) and that your 5+ labels are spread across different time periods. If problems persist, contact support@getpharo.com.
Can I retrain my model?
Yes. As you add more labels over time, you can retrain to produce an updated model that benefits from the additional data. More labels generally lead to more accurate predictions.
Medical
Is Pharo a medical device?
No. Pharo is a general wellness app that helps you notice patterns in your own data. It does not diagnose, treat, or prevent any medical condition. Always consult healthcare professionals for medical decisions.
What should I do if my risk score is elevated?
An elevated risk score means your vitals are showing a pattern similar to what preceded your labeled episodes. Use this as one input among many—pay attention to how you're feeling, take any preventive steps that work for you, and consult your doctor if you have concerns. Pharo helps you notice patterns, not make medical decisions.