Davide Salomoni

Possible Topics

• A Cloud-based architecture and application for multi-disciplinary volunteer computing

In Volunteer Computing, people provide unused computing resources to projects. These resources may range from laptop or personal computers in idle mode, to unused data center capacity. In this thesis, the adaptation of an open-source framework to a Cloud national infrastructure will be studied, so that scientists may define multi-disciplinary computational problems, and users may then subscribe to one or more of these problems, organised as projects, thus contributing to solving them through their unused computing capacity. Topics such as authentication, authorisation, user interfaces and workload/data distribution will be an integral part of this thesis. Once developed as a Proof of Concept, the Cloud-based volunteer computing solution will be deployed over INFN-Cloud and tested with one or more bioinformatics projects.

• Generating actionable and customisable metadata descriptions for the processing of large volumes of structured and unstructured health-related data.

In several applications, it is convenient to retain data originated from multiple sources in its original format, rather than forcing it into a single database technology. However, even with this approach it is necessary to standardise the metadata for the various data sets, so that data analysis can be performed without too many complications. This topic will explore the use of Argo Workflows [https://argoproj.github.io/argo-workflows/] to run CI/CD pipelines on Kubernetes over INFN Cloud, so that the entire process of data ingestion, data quality checking, and metadata creation or harmonisation can be automatised and made easily reproducible. A concrete application of this topic will be performed on datasets currently being collected by the PLANET project [https://indico4.twgrid.org/event/20/contributions/1083/attachments/692/806/PLANET-ISGC22-%20final_Spiga.pdf], related to the analysis of possible correlations between common pathologies and environmental factors.

• Integrating GraphQL with a Redis-based metadata cache for health-related data stored in the Cloud through MinIO.
• Integrating Redis streams in a big data platform for the ingestion of real-time data coming from Internet of Things (IoT) devices.
• Analysis of heterogenous data sets via Deep Learning and Machine Learning models locally served through RedisAI.