Software development teams are confronted with substantial problems when it comes to software flaws, which can result in reduced user experience, compromised security, and lower reliability. For software systems to be robust and mitigated, effective bug detection and patching are essential. The goal of this research is to improve bug detection and repair in software development processes by using an effective way for code representation and then fine-tuning the model on real bugs. The paper explores and highlights the mechanisms, applications, and implications of several bug localization techniques, including machine learning-based approaches and graph based methods. This research adds to the continuing conversation on software quality assurance and reliability by clarifying the nuances of bug localization and patching. This research helps in improving the false positive rate by identifying the real bug correctly. It highlights the graph-based model, which works on finding short distance in graph for patching up the bug, as the best method which can effectively use the real bugs for automatic program repair.