The efficacy of signature-based malware detection is limited in detecting novel forms of malware that are created through obfuscation and deformation methods. AI-powered techniques are extensively utilized to detect malware by employing classification models that analyze its behavioral patterns. This research presents a novel malware detection approach, referred to as PEMA, which integrates multiple effective AI models that are carefully chosen through an automated machine learning framework. Parallel ensemble learning with XGBoost, CatBoost, and LightGBM algorithms is employed by PEMA to make deep malware analysis faster and more accurate. Since PEMA was put together using three separate models, it can lessen the likelihood that it will be susceptible to poisoning and adversarial attacks. The performance of PEMA is verified by experimenting with well-known datasets, such as the EMBER 2017 and 2018, for evaluation purposes. As a result, PEMA attained Fl scores of 99.41% and 97.64%, respectively. Furthermore, it exhibits superior performance compared to other recent methods utilizing the same datasets and improves significant detection time.