An end-to-end NLP pipeline for detecting and mitigating AI-generated text watermarks using transformer models from Hugging Face with ML-based ensemble decisions. It transforms text via paraphrasing, back-translation, synonym replacement, and sentence restructuring, then compares detection results before and after processing.