This seminar focuses on Machine Learning approaches that specialize in sequential data. Most real-world data is acquired over time. Moreover, most of the available data is not image data. We will discuss works before the Transformer era (e.g., RNNs, LSTMs) and highlight their strengths and weaknesses outside the Computer Vision domain. More recently, transformer-based approaches have outperformed earlier methods. We selectively pick works that highlight their strength in knowledge discovery on sequential data. With the strong trend towards powerful multi-modal models, the seminar aims to introduce state-of-the-art methods to produce robust embeddings based on Time Series data.