Causality is crucial for advancing machine learning (ML) beyond correlation-based models, enabling more robust, interpretable, and generalizable systems. Understanding causal relationships helps improve decision-making, interventions, and predictions, especially in the presence of distribution shifts. Traditional ML models often fail in such settings, relying on spurious correlations rather than true causal structures. Causal representation learning, which combines ML with causal inference, offers a solution by capturing underlying causal mechanisms, enhancing model adaptability and robustness across domains. This thesis explores two approaches within causal representation learning to address distribution shifts and improve generalization. It also connects ML and causal discovery, aiming to uncover deeper causal relationships. By integrating ML techniques into causal discovery, we can improve model predictions, design better systems, and build more trustworthy AI. Additionally, causal discovery aids in identifying relevant causal variables, enhancing the effectiveness of causal representation learning in real-world applications like healthcare and biology.