Hybrid artificial intelligence is rapidly advancing, particularly in the domain of combining ontology and machine learning models. However, existing approaches in this field still encounter several limitations. Most current works tend to combine a single ontology model with a specific learning algorithm and often have a strong focus on specific application domains, which can complicate system adaptation and generalization. To address these limitations, we introduce in this paper an objective-driven, hybrid, and modular approach that promotes the integration of multiple machine learning and ontology models. The approach consists of decomposing the studied application into several tasks, each of them using the most appropriate ontological and machine learning models applied to a subset of knowledge and data. Our approach enhances adaptability and flexibility by tailoring artificial intelligence models to specific goals and reasoning requirements, thereby promoting a more effective hybrid artificial intelligence system and enabling the abstraction and reuse of developed solutions in various application domains. The proposed approach is applied in the design of a hybrid artificial intelligence model for the development of a compact all-optical Arithmetic and Logic Unit.