SOURCE-LINKED INTELLIGENCE

Efficient Foundation Model Inference Across the Computing Continuum

CORDIS · observation · Publication date unknown

new class of AI paradigm that is trained on broad data and is adaptable to diverse downstream tasks across multiple modalities, including text, images, audio, video, and even time-series. They enable generative AI with zero-shot and few-shot capabilities, marking a paradigm shift from task-specific AI systems. This is why FMs are the go-to solution and are getting more and more attention over the last years. However, their adoption faces critical challenges such as i) dependence on centralized cloud infrastructures and providers, which in turn threatens digital sovereignty and data security; ii) high energy demands of running large-scale FMs that conflict with sustainability goals; iii) prohibitive costs risk excluding small and medium enterprises (SMEs) from deploying and customizing FMs tailored to their needs; iv) and reliance on remote services introducing latency and single points of failure. To address these challenges, the f-inference project envisions a Resource-driven Computing Continuum (RCC) that enables the hyper-distribution of FMs and their deployment across Cloud, Edge

Read original source ↗ Open in workspace

recordType

award

status

SIGNED

region

EU

value

7984003.75

unit

EUR