GLUCOTWIN is a smart healthcare application designed to improve diabetes management for children aged 3 to 12 years. Managing diabetes in children is a complex and sensitive task because insulin dosage depends on multiple changing health factors, including blood pressure, body temperature, BMI, blood sugar condition, meal timing, and carbohydrate intake. Guardians often face difficulty in making correct insulin-related decisions without medical guidance. To address this challenge, GLUCOTWIN provides an intelligent and structured system that combines mobile application development, machine learning, and doctor-supported review in one platform. The system is developed as a role-based application with three main users: Admin, Guardian, and Doctor. Guardians can register, add child profiles, enter health values, record meal details, and perform sugar-related tests through a simple mobile interface. Doctors can review prediction results and provide professional medical recommendations, while the admin manages doctor approvals, monitors system activity, and controls overall records. This structured workflow improves safety, organization, and role-based access in the application. A key feature of GLUCOTWIN is its machine learning component for insulin dosage prediction. The system uses health-related input values such as blood pressure, temperature, BMI, sugar level, meal timing, and carbohydrate intake to estimate the required insulin dose. The dataset used in the project contained 3000 records and 11 attributes, and multiple machine learning regression models were trained and evaluated. After performance comparison, the Gradient Boosting Regressor was selected as the final model because it achieved the best results, with high accuracy and low prediction error. This trained model was then saved and deployed through Python for real-time insulin prediction. The application is built using Flutter for the mobile frontend and Firebase for real-time data storage and history management. Firebase stores user accounts, child records, meal details, sugar test data, prediction history, and doctor recommendations. This allows all important records to be managed in an organized and accessible way. The system also supports carbohydrate estimation from food details, which is an important factor in insulin calculation and helps guardians enter meal-related data more effectively. Another important strength of GLUCOTWIN is that it does not rely only on automated prediction. After the machine learning model generates the insulin prediction, the result is reviewed by a doctor who can provide guidance and recommendations before final action is taken. This combination of AI-based prediction and doctor supervision makes the system more reliable and safer than a basic calculator or simple record-keeping app. Overall, GLUCOTWIN is a practical and meaningful healthcare support solution that reduces confusion for guardians, improves decision-making, supports doctors with organized patient data, and creates a connected system for child diabetes care. It combines intelligent prediction, role-based system design, real-time data management, and medical review into one user-friendly platform