At Rubixe Technologies, I built and evaluated NLP pipelines for text classification and sentiment analysis on labeled text data, seeing firsthand how annotation consistency affects model performance.
I designed and refined prompts for a Generative AI Q&A application using the OpenAI API and LangChain. I scored outputs for accuracy and hallucination, helping cut query-resolution time by 60%.
I reviewed and validated data across more than five business domains, catching errors and inconsistencies and reducing data preparation time by 30%. When records were ambiguous, I documented my reasoning and escalated unclear cases rather than guessing.
My projects include a retrieval-augmented Q&A assistant and structured evaluations of LLM responses. As a native Telugu speaker fluent in English, I bring bilingual language judgment to annotation and quality review.
