EmoBot: Real-Time Emotion Detection

• Spearheaded and trained a BERT-based transformer model on a custom tweet dataset, achieving 92% ac- curacy in classifying 7 distinct emotions.
• Optimized the training pipeline with L2 regularization and dropout, slashing validation loss by 54% and eliminating model overfitting.
• Deployed the model as a real-time (under 50ms) chatbot on Hugging Face using Streamlit, providing a seamless user interface for emotion analysis.