Multilingual AI Data Annotation & LLM Evaluation
Contributed to AI training data projects involving annotation, evaluation, and quality assurance of text and speech datasets used to train and improve machine learning models. Worked extensively on LLM (Large Language Model) evaluation, assessing AI-generated responses for accuracy, relevance, coherence, and safety. Performed prompt-response analysis and rating, ensuring outputs followed instructions and maintained logical consistency. Identified and flagged issues such as hallucinations, bias, and unsafe or low-quality content, supporting AI model alignment and responsible AI practices. Handled multilingual datasets (English, Bengali, Hindi), including text annotation, transcription, and linguistic validation. Conducted TTS (Text-to-Speech) evaluations focusing on pronunciation, fluency, and naturalness, as well as ASR (Automatic Speech Recognition) validation for transcription accuracy and speaker consistency. Maintained strict adherence to annotation guidelines, consistently delivering high-quality, accurate, and reliable outputs within deadlines. Collaborated on large-scale data pipelines, contributing to the development and improvement of NLP, conversational AI, and speech-based AI systems.