• Solo capstone: an end-to-end app that classifies customer complaints by intent and sentiment, clusters them by
topic, and drafts a suggested reply.
• Trained TF-IDF + Logistic Regression for intent classification and MiniLM sentence embeddings + KMeans (5
clusters) for topic clustering on the Bitext customer-support dataset; integrated pretrained sentiment (RoBERTa)
and reply-generation (Qwen2.5) models.
• Live: nlp-customer-support-analyzer.streamlit.app