Artificial Intelligence is rapidly revolutionizing higher education by increasing personalized learning, promoting student engagement & allowing data-driven academic support. This study evaluate the influence of AI driven tools like ChatGPT, Grammarly, Gemini, and Duolingo on personalized learning among university students in Bangladesh. We analysed a dataset that contained 1,511 student responses to 12 features of student experience with AI-supported learning using both qualitative and quantitative analysis techniques. The analysis included the application of several machine learning models (including Decision Tree, Random Forest, K-Means and Baseline classifiers), which were used to predict student experiences with AI-supported learning. The Random Forest model obtained the highest accuracy of 90%, whereas the Random Uniform classifier showed the lowest accuracy of 33%. Highlighting the different effective of these models. The result that AI enhance understanding, motivation, and academic performance through customize feedback & adaptive learning. Despite challenges like data privacy issues, unequal technological access & certain model limitations. From all the validated models, the highest accuracy was 90%, while the lowest was 33%. The result indicate that AI significantly enhance understanding, motivation & academic performance through personalized feedback & adaptive learning paths. Our study identifies several persistent barriers to successful adoption of AI in higher education in Bangladesh, including the potential for data privacy risk as well as access to technology and the limitations of the models themselves. The results of our research help fill the gap in research on AI adoption in higher education in Bangladesh, while also providing a practical set of recommendations for educators and policymakers to consider in the successful and responsible implementation of AI.