1. Introduction
I’ve been using wearables like Fitbit and Pixel Watch for a while, and I’ve always had the same question in the back of my mind:
“These devices collect so much data… but what does it actually mean for my health?”
I could see my steps, sleep score, heart rate, all the usual numbers — but I didn’t really know how to connect them to real insights about my body. That curiosity is what pulled me into this world of AI for personal health.
As I started reading research papers and experimenting with data, I realized something important:
AI is slowly becoming the “missing link” between raw wearable data and actual, useful guidance.
It can interpret patterns, explain them in simple language, and sometimes even coach you like a personal guide.
This blog series is my attempt to share what I learned — not as a researcher writing a formal report, but as a student trying to make sense of a fast-moving field in a friendly, simple way.
What This Series Will Cover (Based on the Timeline of Publications)
I studied four papers and mapped out how AI in personal health evolved over the past few years.
Here’s the order I’ll follow in this series — from earlier work to the most recent:
Intro Blog (this one) - Why I’m exploring this topic and how the whole series is structured.
Blog 1 (2024) — PH-LLM - A large language model that gives personalized sleep and fitness coaching using wearable data.
Blog 2 (2024) — PHIA - An agentic LLM that writes code, analyzes your data, and turns it into meaningful health insights.
Blog 3 (2025) — IR Prediction + IR Explainer Agent - A combination of machine learning and an LLM-based explainer to estimate metabolic risk.
Blog 4 (2025) — Personal Health Agent (PHA) - A multi-agent system where different AI “roles” collaborate — data analyst, coach, and domain expert.
Blog 5 (My Project, 2025) - My own proof-of-concept that mixes ideas from all four systems to create a unified health agent architecture. My goal is to keep everything straightforward, even if someone is reading about AI for the first time.
Why This Topic Matters (and What I Hope You’ll Learn)
Most of us don’t think much about our health until something feels wrong. But our wearables are quietly tracking us every day — our sleep, heart rate, movement, and even stress signals. That means they’re collecting clues long before we ever feel anything.
AI can turn those clues into something meaningful by noticing patterns we might miss, predicting early risks, explaining things in plain English, and even giving small suggestions that feel personal. To me, this makes health feel more understandable and less like a set of random numbers.
Through this series, I want to share what I learned about these systems in a simple, relatable way. I’ll talk about how wearables work with AI, how different models make sense of health data, why LLMs and multi-agent systems matter, where the ethical issues show up, and how these ideas connect to real projects.
I’m writing as a student — sharing what made sense to me, what confused me at first, and what finally “clicked.” And at the end of the series, I’ll also walk through my own proof-of-concept project where I combine everything I learned into one architecture.
10. References
PH-LLM: A Personal Health LLM for Sleep & Fitness Coaching
Nature Medicine (2025)
https://www.nature.com/articles/s41591-025-03888-0PHIA: Transforming Wearable Data into Health Insights Using LLM Agents
Google Research Blog (2024)
https://research.google/blog/advancing-personal-health-and-wellness-insights-with-ai/SHARP Framework: Principles for Building Health & Wellness LLMs
Google (2025)
https://services.google.com/fh/files/blogs/winslow_2025_sharp_framework.pdfInsulin Resistance Prediction From Wearables + Routine Blood Biomarkers
arXiv (2025)
https://arxiv.org/pdf/2505.03784Personal Health Agent (PHA): Multi-Agent System for Data + Coaching + Expertise
arXiv (2025)
https://arxiv.org/pdf/2508.20148
Top comments (0)