An architecture for a serverless web application with infrastructure as code
Nguyen, Minh Tuan (2025)
Kandidaatintyö
2025
School of Engineering Science, Tietotekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2025072479144
https://urn.fi/URN:NBN:fi-fe2025072479144
Tiivistelmä
Deploying AI-powered applications, particularly those using generative models, remains a complex and resource-intensive task—often requiring GPU-equipped infrastructure and expert operational management. This thesis addresses that challenge by proposing a lightweight, serverless architecture tailored for small teams and individual developers.
The aim of the study is to design and evaluate a cloud-native system that integrates serverless computing, Infrastructure as Code (IaC), and AI-as-a-service to enable scalable, low-cost deployment of generative AI web applications. The architecture consists of a React-based frontend hosted on Azure Static Web Apps, a backend powered by Azure Functions, and image storage managed through Azure Blob Storage. Generative functionality is provided via the Replicate API, while all infrastructure is provisioned using Terraform and deployed via GitHub Actions.
The system was developed using a design science methodology and evaluated through experimental implementation. Results show that the architecture achieves effective separation of concerns, supports reproducible and automated deployment, and enables AI-driven image generation with minimal overhead. Identified limitations include cold start latency, dependency on third-party APIs, and basic secret management.
This work demonstrates that modern cloud platforms can support practical, maintainable, and scalable AI applications without requiring extensive infrastructure expertise—offering a viable solution model for rapid development in educational, experimental, or lightweight production contexts.
The aim of the study is to design and evaluate a cloud-native system that integrates serverless computing, Infrastructure as Code (IaC), and AI-as-a-service to enable scalable, low-cost deployment of generative AI web applications. The architecture consists of a React-based frontend hosted on Azure Static Web Apps, a backend powered by Azure Functions, and image storage managed through Azure Blob Storage. Generative functionality is provided via the Replicate API, while all infrastructure is provisioned using Terraform and deployed via GitHub Actions.
The system was developed using a design science methodology and evaluated through experimental implementation. Results show that the architecture achieves effective separation of concerns, supports reproducible and automated deployment, and enables AI-driven image generation with minimal overhead. Identified limitations include cold start latency, dependency on third-party APIs, and basic secret management.
This work demonstrates that modern cloud platforms can support practical, maintainable, and scalable AI applications without requiring extensive infrastructure expertise—offering a viable solution model for rapid development in educational, experimental, or lightweight production contexts.