Ankit Punjabi

Master's

Data Science

LLM API Call Enhancement

- Recreated the research project "Do Pre-trained Code Models Possess Knowledge of Correct API Names?" from scratch, boosting model accuracy by 15 - Enhanced CodeBERT, GraphCodeBERT, and GPT-3.5 models for API name prediction, achieving a 20% improvement in prediction accuracy - Masked over 50,000 API names within the dataset, improving model generalization and reducing overfitting - Leveraged the INK framework and CSNet models to rigorously evaluate and refine the models, resulting in a 25