This paper is intended as a preliminary analysis for defining and orchestrating workflows in a Cloud Continuum (CC) environment. Starting from an analysis of interactive computing platforms, in particular some modified kernels of Jupyter Notebook for executing code in heterogeneous and distributed environments, it moves towards integration with a source-to-source compiler based on code decomposition and annotation for parallelization and automatic distribution through templates. Finally, integration with tools for defining and executing workflows in heterogeneous environments is discussed. Within the project and in the paper, several different approaches are presented to integrate the two techniques at different levels (from notebook to modified kernel). The aim is to facilitate the definition and execution of workflows for fine-tuning basic artificial intelligence models and large language models.