<![CDATA[Medication non-adherence remains a persistent challenge in long-term treatment, and mobile health applications can support medication management through reminders, education, and self-monitoring. This study designed and implemented “SehatPol!”, an integrated self-monitoring health system using a Modified Waterfall approach. The system consists of a Flutter mobile application for users and a React-based web dashboard for administrators, with Firebase Authentication and Cloud Firestore as Backend-as-a-Service components. Google ML Kit Face Detection is used on-device as an additional interaction step before a medication-intake log can be recorded; this mechanism detects the presence of a face but does not verify user identity or prove medication ingestion. The Google Gemini API is integrated as a conversational agent using context-aware prompting to limit responses to medication-management and educational context; clinical factuality, hallucination rate, and triage accuracy were not evaluated in this study. Functional testing showed that all executed unit and black-box test cases passed. Preliminary usability testing with five users produced a mean System Usability Scale (SUS) score of 68.5 (SD = 15.27; range = 47.5-82.5). A weighted task-completion score averaged 72.5%, with medication scheduling obtaining the lowest score (50%) and longest completion time. An adapted heuristic evaluation by three evaluators yielded a mean conformity score of 3.95/5. These results indicate that SehatPol! is functionally feasible as a medication-management support prototype, while further work is required to validate clinical effectiveness, identity/liveness verification, AI safety, privacy, and usability in larger samples.Keywords— mHealth, Conversational Agents, Medication Adherence Support, Face Detection, System Usability Scale]]>
