At Genpact, I evaluate enterprise conversational AI responses using rubric-based scoring and create gold-standard chosen responses for complex queries. I also audit outputs against regional statutory SOPs and knowledge base documentation to identify hallucinations and support policy alignment.
I built a failure mode taxonomy to inform fine-tuning and prompt engineering feedback loops, and defined intent classification and resolution boundaries for AI agents. As a Data Science Intern at G.H. Raisoni College of Engineering, Nagpur, I built domain-specific text summarization for Indian languages using MT5, IndicBERT, and mBART.
