Multimodal Sentiment Analysis | PyTorch, BERT, ResNet, Torchaudio Built a deep learning model that predicts sentiment (Positive, Neutral, Negative) from text, audio, and facial image inputs using BERT, MFCC + CNN, and ResNet18; implemented multimodal feature fusion and classification with PyTorch and Hugging Face Transformers.