Bipul Bhatta

I've trained and evaluated LLMs at Joinstellar.ai and Outlier, reviewing, ranking, rewriting, and quality-checking responses across coding, reasoning, generalist, and multimodal workflows. I create task-specific rubrics, identify failure modes, and deliver actionable feedback for RLHF and supervised fine-tuning datasets.

At ETAS Ltd, I led end-to-end development of an AI chatbot for querying AUTOSAR specification documents using Python, LangChain, and Azure. I also built retrieval workflows, evaluated domain-specific model responses, and produced compliance findings and technical reports for engineering teams.