Research of architectural approaches to cross-platform mobile application development using Kotlin Multiplatform

Abstract

The paper investigates architectural approaches to the development of cross-platform mobile applications based on the Kotlin Multiplatform (KMM) technology. The current state of mobile development is characterised by a technological duality of two dominant platforms, Android and iOS, which rely on fundamentally different programming languages, frameworks, and platform-specific application programming interfaces. This situation leads to doubled development effort, duplicated defects, and increased product maintenance costs. The evolution of cross-platform solutions is reviewed, including hybrid WebView-based approaches (Apache Cordova, PhoneGap, Ionic), the Xamarin technology with C# and the Mono runtime, and the modern frameworks React Native and Flutter. It is established that each of these approaches entails significant trade-offs in terms of performance, naturalness of user experience, or integration with platform-specific application programming interfaces. The Kotlin Multiplatform technology developed by JetBrains implements a fundamentally different approach: sharing business logic written in Kotlin while preserving fully native user interface layers. The paper proposes a multi-module architecture for KMM applications with a clear separation into presentation, domain, data, and infrastructure layers. The expect/actual mechanism is analysed as the principal instrument for handling platform-specific code. The role of key infrastructure libraries of the Kotlin ecosystem is examined: Ktor for network operations, SQLDelight for database operations, kotlinx.coroutines for asynchronous execution, Koin for dependency injection, and kotlinx.serialization for data serialisation. A comparative evaluation of four approaches (native development, KMM, Flutter, and React Native) is performed using the criteria of cold start time, binary file size, runtime memory consumption, and shared code ratio. The obtained results demonstrate that Kotlin Multiplatform delivers performance close to native, with a cold start time overhead of approximately 13 % and a share of common code reaching 75 % in a typical application. The results substantiate the use of KMM for projects in which user experience quality and performance are critical and the maintenance of two independent codebases is no longer economically justified. The prospects for further research are also analysed, including the adoption of Compose Multiplatform as a unified user interface framework, integration with on-device machine learning, and the broader use of KMM in enterprise environments.