pavithra s

I've built AI-driven projects including a Telecom AI Customer Support Copilot that uses RAG, DistilBERT, FAISS, and Hugging Face Transformers to generate context-aware responses from telecom knowledge bases.

I achieved a 92% F1-score for customer intent classification across five categories and analyzed 20,000 synthetic telecom interactions using exploratory data analysis, sentiment analysis, and NLP. I also automated query categorization and response recommendations through a Streamlit workflow.

For a Driver Drowsiness Detection System, I developed a TensorFlow-based image classification model that achieved 95% accuracy in detecting drowsy and alert driver states. I processed more than 10,000 facial images and improved robustness through augmentation and hyperparameter tuning.

I'm eager to apply Python, SQL, machine learning, deep learning, NLP, and data analytics to create data-driven solutions for real-world business challenges.