Artificial intelligence as-a-service in agriculture: sketching a scalable platform for multipurpose decision support

Precision agriculture (PA) systems generate extensive amounts of data, ideal for training supervised learning models. However, while, these models are often developed independently to address specific problems, mostly focusing on data engineering and AI architectures, creating scalable, cost-effective, and user-friendly AI abstraction layers that adapt to diverse challenges remains underexplored. Currently available cloud-based AI platforms are mostly commercial – e.g. Amazon Web Services (AWS) –, requiring subscriptions that are not affordable for small to medium-scale farmers. Therefore, in this position paper, an AI-as-a-Service (AIaaS) platform supporting multiple AI tasks with modern open-source tools integrable with PA systems is proposed, with the goal of enabling a cost-effective dissemination through farmers community and related professionals. This platform has already some fully functional modules that will be demonstrated on a PA sourced grapevines imagery, encompassing classification and segmentation tasks.