In this chapter, we explore modern emerging trends in machine learning (ML) architecture. We will cover the main advancements like Vision Transformers (ViTs), transformer models beyond natural language processing, the rise of agentic AI, and their applications in multi-modal learning. We also include practical samples such as transformer-based image classification. Additionally, we talk about pre-training, self-supervised learning, and fine-tuning, showcasing their potential for revolutionizing ML workflows. Afterward, we talk about AI in edge computing. Edge devices include your personal computers, cellphones, and IoT devices, and for many reasons, running ML models on them is interesting. By that, users have higher privacy, especially in sensitive data; they will experience less delay and can use their models in places with limited network coverage. However, running models in an edge device is not easy because of the limited resources of the edge device. In this chapter, we will discuss various benefits and challenges of ML in edge devices, and we will provide some suggestions to overcome challenges.