This paper introduces SePA (Search-enhanced Predictive AI Agent), a novel LLM health coaching system that integrates personalized machine learning and retrieval-augmented generation to deliver adaptive, evidence-based guidance. SePA combines: (1) Individualized models predicting daily stress, soreness, and injury risk from wearable sensor data (28 users, 1260 data points); and (2) A retrieval module that grounds LLM-generated feedback in expert-vetted web content to ensure contextual relevance and reliability. Our predictive models, evaluated with rolling-origin cross-validation and group k-fold cross-validation show that personalized models outperform generalized baselines. In a pilot expert study (n=4), SePA's retrieval-based advice was preferred over a non-retrieval baseline, yielding meaningful practical effect (Cliff's δ=0.3, p=0.05). We also quantify latency performance trade-offs between response quality and speed, offering a transparent blueprint for next-generation, trustworthy personal health informatics systems.
@article{arxiv.2509.04752,
title = {SePA: A Search-enhanced Predictive Agent for Personalized Health Coaching},
author = {Melik Ozolcer and Sang Won Bae},
journal= {arXiv preprint arXiv:2509.04752},
year = {2025}
}
Comments
Accepted at IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI'25). 7 pages, 5 figures, 3 tables