DataRiver has received funding for the project “DataRiver AI Platform: Development of a Multimodal Architecture Based on Generative AI, Tabular Foundation Models, and Agentic Workflows for the Integration and Intelligent Querying of Corporate Information Assets,” CUP: E71B26000220009, funded under the “PR FESR 2021-2027” call for proposals – ACTION 1.1.1 – CALL FOR PROPOSALS TO AWARD GRANTS TO COMPANIES FOR RESEARCH AND DEVELOPMENT PROJECTS THROUGH THE RECRUITMENT OF NEW TALENT 2026,” promoted by the Emilia-Romagna Region.
The DataRiver AI Platform project aims at the experimental development of a unified platform based on Generative Artificial Intelligence, Tabular Foundation Models (TFM), and Agentic Workflows. The goal is to overcome the siloed fragmentation of corporate information by enabling the simultaneous querying, in natural language, of structured databases and complex unstructured documents within a single intelligent environment
The research is structured around three Specific Objectives: 1) development of a hybrid RAG architecture with semantic reranking to reduce “hallucinations” in AI models; 2) integration of TFM and Action Agents based on Function Calling to transform queries into operational commands on enterprise management systems, following a “zero-menu” paradigm; 3) design of vertical agents for the Life Sciences domain, for the automated analysis of sources such as PubMed and EMA
The expected outcome is a model-agnostic platform, TRL 6–7, capable of democratizing access to data and accelerating the digital transition of strategic supply chains.