A Conversation with Sorina Pop

We spoke with Sorina Pop, researcher at CREATIS - a joint research centre among CNRS, INSERM, INSA Lyon, Claude Bernard Lyon 1 University, and Jean-Monnet University of Saint-Etienne in France - and coordinator of VIP, Virtual Imaging Platform. 

VIP is a web portal for medical imaging applications. It allows users to access scientific applications as a service directly through a web browser and to utilise distributed computing resources transparently. It leverages resources within the biomed virtual organisation of the EGI e-Infrastructure to provide an open service for researchers worldwide.VIP was launched around 2010, in collaboration with EGI's Biomed virtual organisation, to address key challenges faced by domain experts who lacked computing expertise.

VIP was conceived to provide transparent access to compute and storage resources, similar to modern cloud services but tailored for researchers focused on medical imaging rather than informatics. Early hurdles included a steep learning curve with EGI resources, which VIP mitigated by enabling access-as-a-service for resource-intensive applications such as Monte Carlo simulators. Over the years, it evolved to support open and reproducible science natively, with tools ensuring traceability without burdening users.

The platform primarily leverages HTC services via Dirac for job scheduling and storage, alongside CVMFS for software deployment and EGI Check-in for authentication. Dirac's flexibility enables seamless use of HTC or cloud resources, including customised Virtual Machines for high-memory requirements, while maintaining transparency for users. While cloud interest grows, Dirac enables opportunistic resource use without backend rewrites.

Joint European projects such as EGI-ACE and EOSC Data Commons have strengthened ties, fostering technology adoption and expert exchanges across communities. VIP now serves over 1,500 users worldwide, prioritising steady impact through publications over explosive growth. A secondary private instance deployed in France for projects dealing with sensitive data demonstrates VIP's deployability on local clusters or EGI, supporting both Dirac and direct batch schedulers.

Recent advances standardise application imports for reproducibility, with ongoing code refactoring and EOSC Data Commons integrations expanding APIs for dispatchers like data matchmakers. VIP's domain-agnostic design can suit other disciplines beyond life sciences, and distributed computing via EGI has been pivotal: much early work would have been impossible without it. Emerging AI uses within VIP already include running trained models and GPU training, while agentic planning for experiment design is next in the development pipeline.

EGI's enduring support, reliable infrastructure, and shared values around open science excellence underpin the collaboration's success. Technical reliability persists, and the relationship thrives on mutual benefits, positioning VIP prominently in EOSC roadmaps.