Periodic Reporting for period 2 - AI4EOSC (Artificial Intelligence for the European Open Science Cloud)
Période du rapport: 2023-09-01 au 2025-08-31
The EOSC is transforming scientific research by offering a unified, secure, and open infrastructure for data management, sharing, and processing. Within this innovative ecosystem, the AI4EOSC project stands out as a key initiative dedicated to making advanced AI tools and services widely accessible. By doing so, we empower researchers and practitioners across Europe to leverage the power of machine learning (ML) and deep learning (DL) technologies more effectively. The rapid advancement of AI and ML has opened new horizons for scientific discovery, but it has also presented significant challenges. Researchers often struggle with the complexity of integrating AI tools into their workflows, especially when dealing with distributed datasets.
The AI4EOSC project addresses these challenges by providing feature-rich services that enable the development and deployment of customizable AI applications, driven by the following key objectives:
- Enhancing ML and DL Capabilities: We provide ML practitioners with advanced services that support the building and deployment of customizable ML, DL, and AI applications. This includes enhancing existing cloud services to support ML and DL on distributed datasets, with a particular focus on federated learning. By delivering methods to build and compose ML and DL tools, we enable the development of more complex data-driven composite AI applications.
- Fostering Collaboration and Exchange: AI4EOSC fosters a vibrant ML and DL exchange within the EOSC ecosystem. By enhancing and increasing the application offer currently available in the DEEP Open Catalogue, we create a rich repository of AI tools and services that can be easily accessed and utilized by researchers across Europe.
- Supporting the EOSC Continuum: AI4EOSC supports the EOSC continuum by providing scalable and flexible AI services that can be seamlessly integrated into the existing EOSC infrastructure. This ensures that researchers have access to cutting-edge AI tools and services, regardless of their location or the specific requirements of their projects.
The AI4EOSC project addresses these challenges by providing feature-rich services that enable the development and deployment of customizable AI applications, driven by the following key objectives:
- Enhancing ML and DL Capabilities: We provide ML practitioners with advanced services that support the building and deployment of customizable ML, DL, and AI applications. This includes enhancing existing cloud services to support ML and DL on distributed datasets, with a particular focus on federated learning. By delivering methods to build and compose ML and DL tools, we enable the development of more complex data-driven composite AI applications.
- Fostering Collaboration and Exchange: AI4EOSC fosters a vibrant ML and DL exchange within the EOSC ecosystem. By enhancing and increasing the application offer currently available in the DEEP Open Catalogue, we create a rich repository of AI tools and services that can be easily accessed and utilized by researchers across Europe.
- Supporting the EOSC Continuum: AI4EOSC supports the EOSC continuum by providing scalable and flexible AI services that can be seamlessly integrated into the existing EOSC infrastructure. This ensures that researchers have access to cutting-edge AI tools and services, regardless of their location or the specific requirements of their projects.
The AI4EOSC project has completed its objectives, delivering a feature-rich platform designed to support the development, training, sharing, and deployment of machine learning (ML) and deep learning (DL) models. Our technical and scientific achievements are outlined below:
1. Feature-Rich Platform: We have successfully delivered a comprehensive platform that supports the entire lifecycle of ML and DL models. This includes tools for model development, training, and sharing, enabling researchers to collaborate more effectively. The platform also facilitates model serving over the cloud continuum, leveraging both serverless computing and dedicated resources to ensure scalability and flexibility.
2. Integration with External Services: A key achievement has been the seamless integration of various marketplaces, platforms, and data repositories. For instance, we have integrated with AI4Life, iMagine, and Zenodo. These integrations enhance the platform's capabilities by providing access to a wide range of AI tools, datasets, and computational resources, thereby enriching the research ecosystem.
3. Prototyping an Open Infrastructure for LLM Serving: One of our notable achievements is the prototyping of an open infrastructure for serving large language models (LLMs) with a privacy-first approach. This infrastructure ensures that sensitive data remains secure while enabling the deployment and utilization of advanced language models. This development is crucial for applications requiring high levels of data privacy and security.
These technical and scientific achievements underscore AI4EOSC's commitment to advancing AI capabilities within the EOSC ecosystem. By delivering a robust, integrated, and privacy-focused platform, we empower researchers to conduct more sophisticated and impactful research, driving innovation and discovery across various scientific disciplines.
1. Feature-Rich Platform: We have successfully delivered a comprehensive platform that supports the entire lifecycle of ML and DL models. This includes tools for model development, training, and sharing, enabling researchers to collaborate more effectively. The platform also facilitates model serving over the cloud continuum, leveraging both serverless computing and dedicated resources to ensure scalability and flexibility.
2. Integration with External Services: A key achievement has been the seamless integration of various marketplaces, platforms, and data repositories. For instance, we have integrated with AI4Life, iMagine, and Zenodo. These integrations enhance the platform's capabilities by providing access to a wide range of AI tools, datasets, and computational resources, thereby enriching the research ecosystem.
3. Prototyping an Open Infrastructure for LLM Serving: One of our notable achievements is the prototyping of an open infrastructure for serving large language models (LLMs) with a privacy-first approach. This infrastructure ensures that sensitive data remains secure while enabling the deployment and utilization of advanced language models. This development is crucial for applications requiring high levels of data privacy and security.
These technical and scientific achievements underscore AI4EOSC's commitment to advancing AI capabilities within the EOSC ecosystem. By delivering a robust, integrated, and privacy-focused platform, we empower researchers to conduct more sophisticated and impactful research, driving innovation and discovery across various scientific disciplines.
By providing accessible and advanced AI tools, we empower researchers to tackle complex scientific challenges more effectively. Our pathway to impact is further strengthened through the easy customization of the platform to meet the specific needs of different user communities. This includes:
- Enhanced Research Capabilities: Researchers have access to state-of-the-art AI tools, enabling them to conduct more sophisticated and impactful research.
- Increased Collaboration: The project fosters a collaborative environment where researchers can share and build upon each other's work, leading to accelerated scientific progress.
- Tailored Solutions: By offering easy customization, AI4EOSC ensures that the platform can be adapted to the unique requirements of various scientific disciplines and user communities, making advanced AI tools more accessible and useful.
In summary, AI4EOSC is transforming the landscape of scientific research by providing accessible, advanced, and customizable AI tools and services. Through our innovative approach and commitment to collaboration, we address the identified problems and needs, contributing to the broader goals of the European Open Science Cloud.
- Enhanced Research Capabilities: Researchers have access to state-of-the-art AI tools, enabling them to conduct more sophisticated and impactful research.
- Increased Collaboration: The project fosters a collaborative environment where researchers can share and build upon each other's work, leading to accelerated scientific progress.
- Tailored Solutions: By offering easy customization, AI4EOSC ensures that the platform can be adapted to the unique requirements of various scientific disciplines and user communities, making advanced AI tools more accessible and useful.
In summary, AI4EOSC is transforming the landscape of scientific research by providing accessible, advanced, and customizable AI tools and services. Through our innovative approach and commitment to collaboration, we address the identified problems and needs, contributing to the broader goals of the European Open Science Cloud.