Phishing attacks pose significant risks in the digital landscape, resulting in financial losses and sensitive information breaches. Traditional detection methods often struggle to keep pace with evolving threats, compromising their effectiveness. This study addresses these limitations by developing a robust detection system using a hybrid machine learning approach. We combine random forest, gradient boosting, and logistic regression algorithms to enhance phishing detection accuracy. A labeled dataset of URLs from Kaggle is utilized, with robust feature engineering extracting key attributes for model training. Following the CRISP-DM framework and leveraging Object-Oriented Programming principles, we develop a model that achieves strong performance metrics. The model's accuracy stands at 84%, with precision, recall, and F1-score values of 85%, 86%, and 84%, respectively. Notably, the model demonstrates excellent ability to differentiate between phishing and legitimate URLs, with an ROC AUC score of 91%. These results confirm the model's potential as a reliable phishing detection tool, capable of identifying phishing URLs effectively while minimizing false positives. Our research contributes to the development of more effective phishing detection strategies, ultimately safeguarding users and organizations from economic and reputational harm. By leveraging machine learning, we can develop more robust cybersecurity systems. Our proposed model can be seamlessly integrated into existing security frameworks to improve the detection of phishing threats.