HusTab Health Analysis Pipeline: Leveraging AI-Driven Wearable Data Analytics for Personalized Health Management

Abstract

The proliferation of wearable technology has led to an unprecedented surge in personal health data, yet a significant gap persists in transforming this raw data into actionable, personalized insights for effective health management. Current systems often overwhelm users with data, lack deep personalization, offer impersonal communication, and possess superficial analytical capabilities. This Master’s thesis by Hamdan Abdulla Naser Salem AlShkeili, titled ”HusTab Health Analysis Pipeline: Leveraging AI-Driven Wearable Data Analytics for Personalized Health Management,” addresses these challenges by proposing the HusTab Health Analysis Pipeline. HusTab is a comprehensive, AI-driven, multi-stage data processing architecture designed to ingest diverse personal health metrics from wearables, subject them to rigorous analysis, and deliver highly personalized, contextually relevant, and actionable health intelligence. The system uniquely integrates a dynamic User Memory System for longitudinal learning and contextualization with a multi-agent Large Language Model (LLM) framework for advanced data interpretation, nuanced insight generation, and empathetic user communication. Key pipeline tasks include data ingestion, validation, correlation detection, multi-agent LLM-driven data analysis, LLM-driven distillation synthesis, user memory updates, personalization, and LLM-driven message generation, with an interactive follow-up process to enrich user context. The research aims to develop an integrated framework for real-time health monitoring and proactive communication, and to leverage LLMs for creating engaging, personalized user messages. The proposed HusTab system seeks to empower users with a deeper understanding of their health, facilitate proactive health management, and foster sustained positive behavioral changes by transforming complex data into meaningful guidance. The research explores the system’s efficacy through detailed use cases, demonstrating its adaptability to different health profiles and its potential to advance personalized digital health.

Date of Award	2025
Original language	American English
Supervisor	Rabeb Mizouni (Supervisor)