Artificial intelligence (AI) workloads such as machine learning (ML) model training and inference have unique computational requirements that can benefit greatly from optimization. Owing to the recent surge in the adoption of AI across industries, new challenges related to the management of large-scale ML workloads have emerged. This chapter provides a high-level survey of techniques for optimizing AI applications for productivity, cost efficiency, and performance, without delving into technical intricacies. Commonly used optimization approaches are categorized into five broad dimensions: hardware, software, data, model, and hybrid optimization. Concepts such as hardware acceleration using specialized chips, software frameworks, and libraries for AI, data management techniques, model simplification methods, and synergistic approaches combining multiple strategies are briefly discussed. This overview is designed to offer technology decision-makers a foundation for understanding the contemporary landscape of AI optimization; it can enable them to grasp potential benefits and trade-offs of various techniques for specific use cases without requiring in-depth technical knowledge, ultimately leading to improved efficiency and performance in AI operations.