This talk explores the advancement of physics-guided and mechanism-inspired artificial intelligence (AI) in the field of computational engineering, which integrates engineering knowledge with cutting-edge AI frameworks to enhance the predictive performance, fidelity, and interpretability by means of neural computational methods. The presentation will mainly focus on the development of novel deep neural networks as well as feature learning techniques that are intrinsically and extrinsically inspired by distinct engineering principles, significantly improving our understanding of physical phenomena. It is believed that the audience will gain insights and explore future prospects in harnessing data-driven approaches with domain-specific knowledge to address engineering challenges.