Primary focus on generating synthetic datasets in a few-shot setting for intent and entity classification. Enhancing a model's generalization ability by appending gold standard dataset with a synthetic dataset.Utilizing Llama3 and an entity based seed dataset to generate multiple datasets through various techniques. Each synthetic dataset was evaluated using a fine-tuned BERT model, and the techniques were ranked based on performance. The top-performing synthetic dataset underwent multiple filtration layers to remove mislabeled intent classes, ensuring high-quality data for improved model training.

Contribute to an open source research group aimed at exploring AI-based solutions for digital learning.Develop and test NLP models for neural machine translation, intent recognition, and dialogue management.Create, test and deploy training and inference pipelines on cloud infrastructure for ML lifecycle.

Teach and assist graduate students to learn and implement Python for Information Security course. Design lab coursework and grade assignments.