Machine Learning, NLP, and AI research projects
State-of-the-art summarization methods for Hindi news articles using BART and T5 models. Achieved Rouge-2 score of 21.48 with IndicBART. Deployed as API on Hugging Face Spaces.
ML system to identify customers likely to leave. Ensemble methods (KNN, Random Forest, AdaBoost, XGBoost) achieving 94% recall for at-risk customer identification.
Legal document summarization using transformer models (T5, BART, PEGASUS, LED). Evaluated on BillSum dataset with both abstractive and extractive methods.
Dataset of 3,000+ Indian legal judgments with summaries. Published research on legal document summarization with fine-tuned LED model on Hugging Face.