A machine learning infrastructurefor Aline using Amazon Web Services

With the rise of Internet of Things, cloud computing has become an increasingly important concept. Applications can be run on hardware with limited capacity due to the heavy computing being transferred to data centers over the internet. This has created new opportunities to what mobile applications are able to do and opened up new horizons for tailor-made and user centered solutions for users through machine learning and the use of data integration. In this report we developed an architecture and infrastructure for a mobile application named Aline. Aline is a learning platform that recommendslearning material to its users according to the users learning objectives. The purpose of the infrastructure was to deploy a framework for supplying the recommendation system with data and providing these recommendations for the end user. The architecture was built around AWS and AWS Lambda was used as core of the infrastructure. AWS Lambda controlled the transfer of datasets from an external database to the AWS Data Lake, interacted withthe machine learning environment and handled https requests from the Aline application to the AWS. This infrastructure consumed 3.5-5.5 seconds tomake recommendations. However, we further optimized it to 1.2-2.6 seconds. Furthermore, the system was tested for scaling using 100 000 test users. We also developed a machine learning algorithm that successfully demonstrated the usability of our architecture.The architecture can be further improved by storing datasets in another format or integrating the Data Lake and the database to a common solution