Informal caregivers often have difficulty managing behavioral symptoms that accompany Alzheimer's Disease (AD).1–3 This can lead to increased caregiver stress, AD patient institutionalization, and other negative outcomes.4,5 Advancement of mobile technologies represents significant potential to innovate low-cost/high-impact mobile applications (apps) as interventions. However, despite increasing prevalence and potential, apps have shown mixed results in terms of wide-spread adoption and sustainable use over time.5 Using data specific to individual AD caregivers and patients, machine learning has significant potential to provide timely, responsive, and individualized behavioral management approaches for AD caregivers to use in day-to-day caregiving, potentially improving AD patient outcomes. However, to realize machine learning potential, app design must implement user-centered design to optimize data acquisition. HFE optimizes the user-centered design of systems based on capabilities/limitations of people within the system and has potential to enhance design of machine learning-based app development for accessibility/optimization in home environments. Objective: Use HFE to develop a user-centered app that implements machine learning approaches to harness the potential of the AD caregiving network and create individualized behavior-management strategies for AD caregivers. We employed an iterative user-centered design process guided by HFE including: design persona creation, rapid prototyping, and contextual interviews/usability testing with AD caregivers (n=10), clinical experts in AD care (n=7), and community-based AD care experts (n=9). User-centered design processes generated four components critical to ensuring potential of machine learning-based individualized care: 1) language used to describe behavioral symptoms and management strategies must be created by AD caregivers, tailoring app lexicon to individual caregiving networks; 2) strategy-logging function must provide machine learning-based recommendations grounded in AD caregiving network experiences (e.g., collaborative filtering; machine learning from past strategies); 3) AD caregivers must be able to input feedback regarding how well strategies worked quickly and easily; and 4) feedback collected on how well strategies worked must facilitate app adaptation over time. HFE user-centered design approach facilitated the identification of four distinct components critical to ensuring app machine learning responds to the real-world individual/system limitations in order to ultimately reduce caregiver burden/stress and improve care for AD patients in the community.