This work addresses the dual challenge of aspect-level sentiment classification and topic modeling on product reviews by leveraging fine-tuned transformer models. Traditional sentiment analysis often overlooks the nuanced opinions embedded within specific product features. To overcome this limitation, we propose a method that identifies aspect terms and classifies the sentiment expressed toward each one individually. We fine-tune pre-trained transformer architectures on product review datasets and integrate them with topic modeling to uncover major themes of consumer discourse. Experimental results demonstrate that transformer-based models significantly enhance aspect-level sentiment classification accuracy, while BERTopic generates coherent and interpretable topic clusters. The proposed framework illustrates the effectiveness of deep contextualized representations for extracting fine-grained, actionable insights from user reviews.