SAIFA – Serbian Artificial Intelligence Antenna Factory

Duration: 36 months (2026-2028)

The Serbian Artificial Intelligence Antenna Factory (SAIFA) aims to become a national hub for AI development and testing, supporting the full lifecycle of AI innovation—from model preparation and dataset curation to deployment. Building on national infrastructure (PARADOX and ITE clusters), SAIFA will provide access to computing resources, AI tools, curated datasets, and consulting services for academia, public administration, startups, and industry. It will serve as a one-stop shop that unifies users across sectors through an integrated AI gateway offering centralized access to models, data, training, and support.

SAIFA’s environment is designed to support real-life AI use cases and provide services for key sectors such as culture and language, healthcare and sustainability (energy, manufacturing, environment), as well as to enable integration of domain-specific applications within the infrastructure based on interoperability and cybersecurity measures.

Through collaboration with both public and private stakeholders, SAIFA will deliver tools, reusable workflows, and domain-specific AI solutions. It will integrate seamlessly with existing national HPC infrastructure and the EuroHPC Pharos AI Factory in Greece and the Italian IT4LIA AI Factory, enabling validation, benchmarking, and scalable deployment of AI applications across Europe. To ensure efficient performance, SAIFA will implement containerized computing, Kubernetes-based orchestration, and dynamic resource scheduling. It will also contribute models, datasets, technical documentation, and synthetic data to the Pharos and IT4LIA catalogues, and engage in cross-border training activities.

Ultimately, SAIFA aims to close the gap between AI research and deployment, stimulate cross-sector collaboration, and strengthen Serbia’s role in the European AI landscape by fostering inclusive, responsible, and sustainable AI innovation.

 

Контакт: Valentina.Janev@pupin.rs

EuroHPC JU, https://www.eurohpc-ju.europa.eu/index_en